National Program on Technology Enhanced Learning

(NPTEL)

Plantwide Control of Integrated Chemical Processes

Indian Institute of Technology Kanpur 208016

Course Coordinator
Dr Nitin Kaistha
Professor
Department of Chemical Engineering
Indian Institute of Technology Kanpur, INDIA
 TABLE OF CONTENTS
Chapter 0 Introduction to Course 01-03
0.1 Background and Motivation 01
0.2 Organization of the Course 03

MODULE I: Essential Process Control Basics

Chapter 1 Process Dynamics 05-11
1.1 Standard Input Changes 06
1.2 Basic Response Types 06
1.2.1 First Order Lag 06
1.2.2 Higher Order Lags 07
1.2.3 Second Order Response 07
1.2.4 Other Common Response Types 09
1.2.5 Unstable Systems 10
1.3 Combination of Basic Reponses 10

Chapter 2 Feedback Control 12-21

2.1 The Feedback Loop and its Components 12
2.2 PID Control 13
2.2.1 The Control Algorithm 13
2.2.2 Controller Tuning 14
2.3 Process Identification 17
2.3.1 Process Reaction Curve Fitting 17
2.3.2 Autotuning 19
2.4 Controller Modes and Action 20
2.5 Rules of Thumb for Controller Tuning 20
2.5.1 Flow Loops 20
2.5.2 Level Loops 21
2.5.2 Pressure Loops 21
2.5.3 Temperature Loops 21
2.5.4 Quality Loops 21

Chapter 3 Advanced Control Structures 23-29

3.1 Ratio Control 23
3.2 Cascade Control 24
3.3 Feedforward Control 26
3.4 Override Control 28
3.5 Valve Positioning (Optimizing) Control 29

Chapter 4 Multivariable Systems 31-39

4.1 Interaction Metrics 32

i
 4.1.1 Niederlinski Index 32
4.1.2 Relative Gain Array 33
4.2 Multivariable Decentralized Control 34
4.2.1 Detuning Multivariable Decentralized Controllers 34

MODULE II: Control of Common Unit Operations

Chapter 5 Control of Distillation Systems 41-64
5.1 Distillation Basics 41
5.1.1 The Simple Distillation Column 41
5.1.2 Splits in a Simple Distillation Column 42
5.2 Basic Control Structures 43
5.2.1 The Energy Balance (LQ) Structure 43
5.2.2 Material Balance Structures 45
5.2.3 Other Control Structure Variants 45
5.3 Temperature Based Inferential Control 45
5.3.1 Single-Ended Temperature Control 47
5.3.2 Dual-Ended Temperature Control 47
5.4 Temperature Sensor Location Selection 47
5.4.1 Maximum Slope Criterion 47
5.4.2 Maximum Sensitivity Criterion 51
5.4.3 SVD Criterion 51
5.5 Considerations in Temperature Inferential Control 52
5.5.1 Effect of LLK/HHK 52
5.5.2 Flat Temperature Profiles 52
5.5.3 Easy Separations 52
5.6 Control of Complex Column Configurations 53
5.6.1 Side Draw Columns 53
5.6.2 Side Rectifier / Side Stripper Columns 53
5.7 Control of Heat Integrated Columns 58
5.8 Control of Homogenous Extractive Distillation System 60
5.9 Plantwide Considerations 61

Chapter 6 Control of Reactors 65-81

6.1 Basic Reactor Types 66
6.2 Plug Flow Reactor 67
6.2.1 PFR Basics 67
6.2.2 Control of PFRs 69
6.2.2.1 Adiabatic PFR 69
6.2.2.1 Cooled Tubular Reactors 70
6.2.3 Intermediate Cooling and Cold Shot Cooled Reactors 72
6.3 Continuous Stirred tank Reactor 73
6.3.1 Jacket Cooled CSTR 75
6.3.2 Reaction Heat Removal as Steam 76

ii
 6.3.3 External Heat Exchanger 77
6.3.4 Cooling Coils 78
6.3.5 External Cooling by Content Recirculation 78
6.3.6 Boiling CSTR with External Condenser 78
6.3.7 Reactor Heat Removal Capacity Constraint 80

Chapter 7 Heat Exchanger Control 82-85

7.1 Control of Utility Heat Exchangers 82
7.2 Control of Process-to-Process Heat Exchangers 84

Chapter 8 Control of Miscellaneous Systems 86-93

8.1 Furnace Controls 86
8.2 Compressor Controls 87
8.3 Decanter Control 89
8.4 Control of Refrigeration Systems 91
8.4.1 Vapor Compression Cycle 91
8.4.2 Vapor Absorption Cycle 91
8.5 Control of Steam Utility System 93

MODULE III: Issues in Plantwide Control System Design

Chapter 9 Control and Steady State Degrees of Freedom 95-107
9.1 Control Degrees of Freedom 95
9.2 Steady State Degrees of Freedom 96
9.3 Degrees of Freedom, Controlled Variables (CVs) and Control Structures 97
9.4 Control Objectives and Choice of CVs 103
9.5 Illustration of Control Objectives and CVs for Example Processes 105
9.6 Snowball Effect 106

Chapter 10 The Pairing Issue: Selection of MVs for CVs 109-116

10.1 Conventional Pairing Approach 109
10.2 Luyben Pairing Approach 109
10.3 Regulatory Control Structure Synthesis Examples: 110
Conventional vs Luyben’s Approach
10.3.1 Single Column Recycle Process 110
10.3.2 Two Column Recycle Process 113

Chapter 11 Economic Considerations in Plantwide Control 117-123

11.1 Economic Process Operation Considerations 117
11.2 Process Operation Modes 118
11.3 Process Constraints and Economic Operation 118
11.4 Approaches for Handling Capacity Constraints 119

iii
 11.4.1 Backed-off Operation 119
11.4.2 Use of Valve Positioning (Optimizing) Controller 119
11.4.3 Altering Material Balance Control Structure Using Overrides 119
11.4.4 Using Constraint Variable as Throughput Manipulator 120

Chapter 12 Economic Plantwide Control Examples 124-133

12.1 Single Column Recycle Process 124
12.2 Two Column Recycle Process 129

MODULE IV: Economic Plantwide Control Design Procedure and Case Studies
Chapter 13 Systematic Economic Plantwide Control Design Procedure 135-144
13.1 Degrees of Freedom (DOFs) and Plantwide Control Structures 136
13.2 Two-Tier Control System Design Framework 137
13.3 Active Constraint Regions for a Wide Throughput Range 139
13.4 Systematic Control System Design Procedure 140
13.4.1 Step 0: Obtain Active Constraint Regions for the 141
Wide Throughput Range
13.4.2 Step 1: Pair Loops for Tight Maximum Throughput 141
Economic CV Control
13.4.3 Step 2: Design the Inventory (Regulatory) Control System 142
13.4.4 Step 3: Design Loops for Additional Economic CV Control 143
at Lower Throughputs Along with Throughput
Manipulation Strategy
13.4.5 Step 4: Modify Structure for Better Robustness / 144
Operator Acceptance

Chapter 14 Economic Plantwide Control of Recycle Process with Side Reaction 145-154
14.1 Process Description 145
14.2 Economic Plantwide Control System Design 146
14.2.1 Step 0: Active Constraint Regions and Economic Operation 147
14.2.2 Step 1: Loops for Maximum Throughput Economic CV Control 147
14.2.3 Step 2: Inventory (Regulatory) Control System 149
14.2.4 Step 3: Additional Economic CV Control Loops and Throughput 149
Manipulation
14.2.5 Step 4: Modifications for a More Conventional Inventory Control 151
System

Chapter 15 Economic Plantwide Control of Ethyl Benzene Process 155-162

15.1 Process Description 155
15.2 Economic Plantwide Control System Design 155
15.2.1 Step 0: Active Constraint Regions and Optimal Operation 155
15.2.2 Step 1: Loops for Tight Control of Full Active Constraint Set 156
15.2.3 Step 2: Inventory (Regulatory) Control System 158
15.2.4 Step 3: Additional Economic CV Loops and Throughput 158

iv
 Manipulation
15.2.5 Step 4: Modifications a for More Conventional Inventory Control 158
System

Chapter 16 Comprehensive Case Study I: Cumene Process 163-183

16.1 Process Description 163
16.2 Economic Plantwide Control System (CS1) Design 164
16.2.1 Step 0: Optimal Steady State Process Operation 164
16.2.2 Step 1: Loops for Tight Economic CV Control 168
16.2.3 Step 2: Inventory Control System Design 170
16.2.4 Step 3: Throughput Manipulation and Additional Economic Loops 171
16.3 Conventional Plantwide Control Structure (CS2) 172
16.4 Dynamic Simulation Results and Discussion 173
16.4.1 Controller Tuning 173
16.4.2 Closed Loop Results 174
16.4.3 Quantitative Dynamic and Economic Comparison of CS1 and CS2 180
16.4.4 Discussion 182
16.5 Conclusions 183

Chapter 17 Comprehensive Case Study II: C4 isomerization Process 184-200

17.1 Process Description 184
17.2 Economic Plantwide Control System (CS1) Design 185
17.2.1 Step 0: Active Constraint Regions and Economic Operation 185
17.2.2 Step 1: Loops for Tight Economic CV Control 187
17.2.3 Step 2: Inventory Control System 187
17.2.4 Step 3: Throughput Manipulation and Additional Economic Loops 188
17.3 Conventional Plantwide Control Structure (CS2) 189
17.4 Dynamic Simulations and Closed Loop Results 193
17.4.1 Tuning of Controllers 193
17.4.2 Closed Loop Results 195
17.4.3 Quantitative Economic Performance Comparison 199
17.5 Conclusions 199

Bibliography 201-202

v
 Chapter 0. Introduction to Course

0.1. Background and Motivation

Chemical processes are designed and operated for manufacturing value added
chemicals, the value addition providing the economic incentive for the existence of the
process. The fiercely competitive business environment constantly drives research and
innovation for significantly improving existing process technologies and for developing new
technologies to satisfy man’s ever growing needs. On the operation side, the processes are
operated to meet key production objectives that include process safety, product specifications
(production rate and quality) and environmental regulations. These key production objectives
must be satisfied even as the process is subjected to disturbances such as changes in the fresh
feed composition, variation in the ambient temperature, equipment fouling, sensor noise /
bias etc. In other words, the process operation must ensure proper management of the process
variability so that the key production objectives are met even in the presence of the process
variability. This naturally leads to the idea of proper management of process variability, the
task accomplished by a well designed automatic process control system.
Consider the heat exchanger in Figure 0.1. Steam is used to heat a process stream to a
certain temperature. Due to variations in the process stream flow rate and inlet temperature,
the stream outlet temperature varies over a large range. From the process operation
perspective, this is unacceptable since the large variation in the process stream temperature
disturbs the downstream unit (eg. a reactor). The installation of a temperature controller that
manipulates the steam flow rate to hold the outlet stream temperature constant mitigates this
problem to a very large extent. This is illustrated in the outlet stream temperature and steam
flow rate profiles in Figure 0.2. For open loop operation (no temperature control), the
temperature varies over a large range while the steam flow rate remains constant. On the
other hand, for closed loop operation (with temperature control), the variation in the outlet
stream temperature is significantly lower with the steam flow rate showing large variability.
The temperature controller thus transforms the variability in the outlet stream temperature to
the steam flow rate. This simple example illustrates the action of a control loop as an agent
for transformation of process variability.

Temperature Controller

TC
Control
Steam in

Valve

TT Transmitter

Process Stream In
HEAT EXCHANGER
Process Stream Out

Condensate Out

Figure 0.1. Heat exchanger with temperature controller

 Figure 0.2. The manipulated (steam flow rate) and controlled variable
(outlet temperature of process stream) with and without control

A chemical process consists of various interconnected units with material and energy
recycle. Controlling a process variable by adjusting the flow rate of a process stream
necessarily disturbs the downstream / upstream process due to the interconnection. Material
and energy recycle can cause the variability to be propagated through the entire plant.
Considering the plant-wide propagation / transformation of process variability, the choice of
the variables that are controlled (held at / close to their set-points), the corresponding
variables that are manipulated and the degree of tightness of control (loose / tight control) are
then key determinants of safe and stable process operation. The choice of the controlled and
manipulated variables is also sometimes referred to as the control structure. Modern control
text books provide very little guidance to the practicing engineer on the key issue of control
structure selection for individual unit operations and the complete process, choosing instead
to focus on the control algorithms and their properties with typical mathematical elegance.
How does one go about choosing the most appropriate plant-wide control structure for a
given set of production objectives? This work attempts to provide an engineering common
sense approach to the practicing engineer for answering this key question.
Given a control system that ensures safe and stable process operation in the face of
ever present disturbances, crucial economic variables must be maintained to ensure
economically efficient or optimum process operation. Depending on the prevailing economic
circumstances, optimality may require process operation at the maximum achievable
throughput or lower throughputs. At the optimum steady state, multiple process constraints
are usually active such as reactor operation at maximum cooling
duty/level/temperature/pressure or column operation at its flooding level etc. How close can
the process operate to these constraint limits is intimately tied with the basic plant-wide
control strategy implemented. The converse problem is that of designing the regulatory
plantwide control system such that the back-off from the constraint limits is the least

1
possible. In this work, we also develop a systematic procedure for designing such an
economic plantwide control system.
In summary, this book is targeted at the practicing engineer to help him design
effective plant-wide control systems through an appreciation of the major issues the control
system must address. It is hoped that the targeted audience finds the work of practical utility.
The author invites suggestions, comments and criticisms for improving upon the work.

0.2. Organization of the Course

The course is organized into four modules, excluding this Introduction. In Module 1, the
reader is introduced to the essentials of process control. Only the most practical aspects of
process control theory are presented. Mathematical rigor is deliberately done away with in
favour of a more colloquial style to keep the discussion focused on the most essential and
practical aspects of process control theory. Module 2 is devoted to the control of common
unit operations found in the process industry. The control of distillation columns is
exhaustively dealt with and includes simple, heat integrated and complex column
configurations. The control of industrially common reactor configurations and heat
exchangers is covered next. Finally common control configurations for miscellaneous
systems such as furnaces, heat refrigeration systems and compressors are discussed. Module
3 elaborates upon the key issues in the design of a plant-wide control system. The need for
balancing the reactant inventory and the interaction between the reaction and separation
section of a plant are described. We then go about developing a systematic procedure for
economic plantwide control system design. Three elaborate case studies on realistic chemical
processes are then presented to demonstrate the application of the methodology. Comparisons
with conventional plantwide control structures show that an economic plantwide control
structure can significantly improve (2-20%) the achievable profit (or maximum throughput)
for a given process. Proper design of the plantwide control system is thus shown to be crucial
to achieving economically optimal process operation.

2
 MODULE I

ESSENTIAL PROCESS CONTROL BASICS

In this module, we cover essential aspects of process control theory, necessary for proper
control system design. A hands-on approach to covering process dynamics, PID control
algorithm, identification, tuning, advanced control structures and multivariable decentralized
control is used, in contrast to the mathematically elegant but abstruse treatment in most
controls texts. Only the most essential and relevant aspects are covered. In the interest of
brevity, since this is a course on plantwide control and not control theory, we do not provide
many detailed solved examples to back the theory and refer the reader to standard text-books
for the same.

3
Chapter 1. Process Dynamics
Process dynamics refers to the time trajectory of a variable in response to a change in
an input to the process. All of us have an inherent appreciation of process dynamics in the
sense that the effect of a cause takes time to manifest itself. It thus takes 20 minutes for a pot
of rice to cook over a flame, 5-10 minutes for the water in the geyser to heat up sufficiently,
years and years of dedicated practice to become an adept musician (or a good engineer, for
that matter!) and so on so forth. In each of these examples, a change in the causal variable
(flame, electric heating or dedicated practice) results in a change over time in the effected
variable (degree of “cookedness” of rice, geyser water temperature or a musician’s
virtuosity). Process dynamics deals with the systematic characterization of the time response
of the effected variable to a change in the causal variable. In process control parlance, the
causal variable is referred to as an input variable and the effected variable is referred to as an
output variable.
In order to fix ideas in the context of chemical processes, Figure 1.1 shows the
schematic of a simple distillation column. An equimolar ABC feed is separated to recover
nearly pure A as the distillate with the bottoms being a BC mixture with trace amounts of A.
The fresh feed, reflux and reboil constitute the inputs to the column while the distillate and
bottoms flow / composition and the tray composition / temperature profiles constitute the
outputs.

F : Feed
D : Distillate
L : Reflux
Vs : Boil-up
B : Bottoms
Q : Heat Duty Pure A
L D
F
A, B & C
(equimolar)

VS

B&C
B

Figure 1.1. Schematic of a distillation column

4
1.1. Standard Input Changes

To systematically characterize the transient response of an output to a change in the

input, the input change is usually standardized to a step change, a pulse change or an impulse
change. These standard input changes are depicted in Figure 1.2. A step change in the input,
the simplest input change pattern, is used in this work to characterize the process dynamics.

Step change Pulse change Impulse change

Area =1
u(t)

u(t)
u(t)

t t t

Figure 1.2. Standard input changes

1.2. Basic Response Types

The dynamics of every process are. Even so, the variety of transient responses can be
characterized as an appropriate combination of one or more basic response types. These
transient responses correspond to the solution of linear ordinary differential equations
(ODEs). Linear ODEs can be compactly represented using Laplace transforms. For example
consider a second order differential equation
d 2 y (t ) dy (t )
τ2 2
+ 2ζτ + y (t ) = K P u (t )
dt dt
where y(t) and u(t) are the process output and input respectively. The Laplace transform
representation in the s domain is obtained by replacing the nth order derivative operator by sn
so that for the second order ODE above
τ 2 .s 2 y ( s ) + 2ζτ .s. y ( s ) + y ( s ) = K P u ( s )
Rearranging, the input-output transfer function becomes
y(s) K
GP = = 2 2 P
u ( s ) τ s + 2ζτs + 1
The ODEs and corresponding Laplace transform representation is noted in Table 1.1.

1.2.1. First Order Lag

The first order lag is the simplest transient response where the output immediately
responds to a step change in the input (see Figure 1.3(a)). The ratio of the change in the
output to the change in the input is referred to as the process gain, Kp. The time it takes for
the output to reach 63.2% of its final value corresponds to the first order time constant τp.
The output reaches ~95% of its final value in 3 time constants.

5
 Table 1.1. Various differential equations and their Laplace transform
Laplace y(s)
Terminology Differential equation
Transform u(s)
Gain y (t ) = K .u (t ) K
du (t )
Derivative y (t ) = s
dt
t
1
Integrator y (t ) = ∫ u (t ).dt
0 s
dy (t ) 1
First order lag τ + y (t ) = u (t )
dt τ s +1
du (t )
First-order lead τ + u (t ) = y (t ) τ s +1
dt
Second Order Lag
Underdamped 2
2 d y (t ) dy (t ) Kp
ζ <1
τ + 2ζτ + y (t ) = K p u (t ) 2 2
dt 2
dt τ s + 2ζτ s + 1
2 Kp
Critically damped 2 d y (t ) dy (t )
τ + 2 τ + y (t ) = K u (t ) 2
ζ =1 dt 2 dt
p
(τ s + 1)
d 2 y (t ) dy (t ) Kp
Overdamped τ 1τ 2 2
+ ζ (τ 1 + τ 2 ) + y (t ) = K p u (t )
ζ >1 dt dt (τ 1s + 1)(τ 2 s + 1)
Deadtime y (t ) = u (t − θ ) e −θ s
dy (t ) du (t ) τ 1s + 1
Lead-lag τ2 + y (t ) = τ 1 + u (t )
dt dt τ 2s +1

1.2.2. Higher Order Lags

If the output from a first order lag is input to another first order lag, the latter’s output
behaves as a second order lag with respect to the input to the first lag. The overall transient
response is S shaped with the output not responding immediately to a change in the input.
When the time constant of the two lags are different, the response is called an over-damped
second order response. The response for the special case where the two time constants are
equal is called the critically damped second order response. Higher order systems result as
more first order lags are connected in series with the transient response becoming
increasingly sluggish.

1.2.3. Second Order Response

Sometimes, a step change in the input causes the output to oscillate before settling at
the final steady state. The simplest such response corresponds to a second order underdamped
system. The damping coefficient, ζ, can be used to characterize all second order responses –
overdamped (ζ > 1), critcally damped (ζ = 1) and underdamped (ζ < 1). The second order
response is shown in Figure 1.3(b).
To gain an appreciation of the impact of damping coefficient on the transient
response, Table 1.2 reports the ratio of the second overshoot to the first overshoot for

6
different values of ζ. A quarter decay ratio is observed for a damping coefficient of 0.218.
Sustained oscillations (decay ratio = 1) are observed for a damping coefficient of 0. As ζ
increases to 1, the overshoot in the output disappears.

Figure 1.3. Output response for unit step change to (a) First order & (b) Second order
process.

Table 1.2. Decay ratio for various different damping coefficients

Damping
Coefficient, ζ
0 0.05 0.1 0.2 0.218 0.4 0.6 1
Decay ratio 1.000 0.730 0.532 0.277 0.250 0.064 0.009 0.000

7
1.2.4. Other Common Response Types

Other types of responses include the pure integrator, the pure dead-time, and the
inverse response. The transient response to a unit step change can be seen in Figure 1.4 and
are self explanatory.

Figure 1.4. The output response for a unit step change for (a) pure integrator, (b)
inverse response and (c) pure dead time process.

8
The most common example of a pure integrator is the response of the tank level to change in
the inlet / outlet feed rate. Unless the inlet and outlet flows are perfectly equal, the tank level
is either rising or falling in direct proportion to the mismatch in the flows. The level in a tank
is thus non-self regulating with respect to the connected flows. A controller must be used to
stabilize all such non-self regulating process variables. Dead time is very common in
chemical processing systems and is due to transportation delay. A very common example of
the inverse response is the response of the liquid level in a boiler to a change in the heating
duty. As the heating duty is increased, the vapour volume entrapped in the liquid increases
causing the liquid interface level to rise initially. Over longer duration, the level of course
reduces since more liquid is being vaporized. As will be seen later, dead time and inverse
response can create control difficulties.

1.2.5. Unstable Systems

Some systems may be inherently unstable. Unstable transient responses are shown in
Figure 1.5. The unstable response may be non-oscillatory or oscillatory as in the Figure.
Reactor temperature runaway is an example of an unstable process. A control system must be
used to stabilize an inherently unstable system.

(a) (b)
Y(t)

Oscillatory Non-oscillatory

t t
Figure 1.5. The output response for unstable process. (a) Oscillatory and (b) non-oscillatory

1.3. Combination of Basic Responses

Any transient response can be reasonably represented as a combination of the above

basic response types. One such combination is the first order lag plus dead time that has been
found to represent the transient response of many chemical processing systems very well. The
response is illustrated in Figure 1.6(a). Another example of such a combination is the inverse
response which can be represented by the parallel combination of two first order lags. One of
the lags has a small gain and a small time constant (ie a fast response) while the other lag has
a gain of larger magnitude and opposite sign with a much larger time constant (i.e. a slow
response in the opposite direction). Figure 1.6(b) illustrates this concept.

9
 (a)

U(s)
KP y(s)
e−θ s
τ ps +1

(b)

Ka
U(s)
τ 1s + 1 + y(s)

—
Kb
τ 2s +1

Figure 1.6. Unit step responses (a) first order plus dead time process (FOPDT) and
(b) Inverse response

10
 Chapter 2. Feedback Control

The safe and stable operation of a process requires that key variables be maintained at
or close to their design values in the face of disturbances entering the process. For example, it
may be necessary to hold a process stream flow rate nearly constant even as the upstream /
downstream pressure fluctuates. Similarly the temperature at the inlet to a packed bed reactor
must be maintained at its design value to prevent reactor run-away and also ensure the
desired conversion to products(s) for varying flow rates of the process stream. Maintaining a
process variable at or near a certain value requires a manipulation handle that can be
appropriately adjusted. For example, the valve opening can be adjusted to maintain the flow
rate through the pipe. Similarly the heating duty of the furnace can be used to heat the process
to maintain the reactor inlet stream temperature. This leads to the idea of feedback control
where the deviation in the variable to be maintained at / near its design value is used to make
appropriate adjustments in the manipulation handle. The variable to be maintained at its
design value is referred to as the controlled variable and the adjustment handle is called the
manipulated variable. The algorithm / procedure used to quantitatively translate the deviation
in the controlled variable to the adjustment in the manipulated variable is known as the
control algorithm.

2.1. The Feedback Loop and its Components

A feedback control loop is schematically illustrated in Figure 2.1. Its primary

components are the sensor, transducer, transmitter, controller, I/P converter and the final
control element. The sensor is the sensing element used to measure the controlled variable
(and other important process variables that may not be controlled). Flow, temperature and
pressure sensors are routinely used in the process industry. Composition analyzers are used
less frequently to measure only key compositions such as the product purity. Most sensors
translate a change in the state of the variable to be measured into an equivalent mechanical
signal such as the stretching / bending of a Bourdon tube. The mechanical signal needs to be
converted into an electrical signal for onward transmission to the control room (or stand-
alone controller). This is accomplished by the transducer. For standardization across different
manufacturers, the range of the input and output signal from a controller is 4-20 mA. The
range corresponds to the sensor / final control element span. The transmitter converts the
electrical signal from the transducer to the 4-20 mA range. The transmitter signal is input to
the controller. The desired value for the controlled variable, referred to as the set-point, is
also input to the controller. The controller output signal is again between 4-20 mA. In the
process industry, this electrical signal is converted to an equivalent 3-15 psig pneumatic
pressure signal using an I/P converter. The pressure signal (or rather change in the pressure
signal) is used to move the final control element to bring about a change in the manipulated
variable. In the process industry, almost all final control elements are control valves that
adjust the flow rate of a material stream.
The controller subtracts the current value of the controlled variable from its set-point
to obtain the error signal as
et = ySP - yt
where y is the controlled variable. The subscript t refers to the current time. The error signal
is input to the control algorithm to determine the change in the manipulated variable (control
input) to be implemented. This is schematically illustrated in Figure 2.2. The most popular
control algorithm, namely the PID algorithm is discussed next.

11
 Set point,
4-20 mA

4-20 mA 4-20 mA
Transducer Transmitter Controller I/P Converter

Pressure, 3-15 psig

Sensor
Control valve (final control element)
PROCESS
Output Input
Variables Variables

Figure 2.1. Schematic of a process with feed back control scheme

e= ySP- yt

ySP e Control valve

Controller Process
(Final control element)

yt
Measuring
element

Figure 2.2. Block diagram of a feed back control

2.2. PID Control

2.2.1. The Control Algorithm

Almost all controllers in the process industry use the Proportional Integral Derivative
(PID) control algorithm. Even as instrumentation and computation technologies have
witnessed a transition from the analog era to the digital revolution, the good old PID control
algorithm remains the most widely used algorithm, not withstanding the onslaught of
advanced model predictive control algorithms. The positional form of the algorithm states
that
 1
t
de 
u t = K C  et + ∫ e(t )dt + τ D t  + bias

 τI 0 dt 
where u is the controller output (input to the process), e is the error in the controlled variable,
and KC, τI and τD are controller tuning parameters. The tuning parameters are referred to
respectively as the controller gain, reset (or integral time) and derivative time. The bias term
in the expression is provided to make the LHS equal the RHS at time t = 0 for proper
initialisation. The three terms in the algorithm correspond to Proportional, Integral and
Derivative action, hence the acronym PID.

12
 The velocity form of the algorithm is more amenable to understanding the effect of
each of the P, I and D actions. Differentiating the above equation, we get
du t  de 1 d 2 et 
= K C  t + et + τ D 
dt  dt τ I dt 2 
The controller gain or proportional gain, KC, determines the fastness of response with larger
values resulting in a fast response to deviations from set-point. This can be verified from the
first term in the velocity form equation where the rate of change of the control input is
directly proportional to the rate of change in the error, KC being the proportionality constant.
The larger the KC, the larger the change in the control input, the faster the return to set-point.
The integral action is provided to ensure zero offset in the controlled variable. If the
controlled variable deviates from its set-point, the controller acts to settle the system at a new
steady state. At this new steady state all time derivatives are zero (by definition) implying the
LHS in the equation above is zero. The RHS also therefore must be zero which requires that
the error term, et, must be zero at the final steady state (t → ∞). The error term in the velocity
form above is due to the integral mode so that integral action moves the control input until
the error in the controlled variable is driven to zero i.e. ensures a zero offset. P and D action
do not guarantee zero offset as at the final steady state, the LHS and RHS terms
corresponding to P and D action are zero. For a P or PD controller with no integral action, the
velocity form of the algorithm imposes no restriction on the output error at the final steady
state. A non-zero offset thus can and does result sans integral action.
The derivative action causes the controller to “think ahead” and is usually introduced
to suppress oscillations from the “seeking behaviour” caused by integral action. In effect, the
derivative action puts brakes on the control action as the controlled variable approaches the
set-point thus avoiding large oscillations around the set-point. Most controllers in the industry
are P or PI controllers and the D action is set to zero. This is because the D action amplifies
noise so that the controller input signal must be pre-filtered appropriately to reap the benefit
of D action. It is easier to simply turn the D action off and properly tune the controller gain
and reset time for the desired control performance.

2.2.2. Controller Tuning

Empirical rules have been developed for tuning PID controllers. These tuning rules
are based on the idea of ultimate gain and ultimate period. Figure 2.3 plots the closed loop
response for a unit step change in the set-point of a first order plus dead time process for a P
only controller as the controller gain is increased. Notice that as the controller gain is
increased, the steady state offset reduces. Also, the response becomes faster. For larger gains
the closed loop response is oscillatory. As the gain is increased further, sustained oscillations
result. Any further increase in the controller gain results in an unstable system with the
oscillations increasing in magnitude with time. The controller gain for which the closed loop
response exhibits sustained oscillations corresponds to the transition from a stable to an
unstable closed loop response. This controller gain at which the closed loop system borders
on instability is referred to as the ultimate gain, KU. The period of the sustained oscillations is
known as the ultimate period, PU. The empirical tuning rules recommend the controller gain
to be a fraction of the ultimate gain and the reset time and derivative time as fractions
(multiples) of PU. Two popular tuning rules are the Zeigler-Nichols and Tyreus-Luyben
tuning rules are tabulated in Table 2.1. For a given ultimate gain and ultimate period, the
controller gain is the least for a PI controller. This is due to the “seeking behaviour” caused
by integral action for zero offset. The closed loop system thus goes unstable for a lower
controller gain implying that it should be lower. The controller gain is the maximum for a

13
PID controller due to the stabilizing effect of D action. As discussed before, D action is
however used rarely in practice due to noise amplification. The PI algorithm is most
commonly used in the industry. The tuning rules show that Zeigler-Nichols tuning is more
aggressive than the Tyreus-Luyben tuning. Application of the ZN tuning rule can cause
process upsets such as a distillation column flooding due to a sudden large increase in the
vapour boil-up caused by a controller. The more conservative TL tuning rule is preferred in
the process industry for a smooth and bumpless handling of transients avoiding large and
sudden changes in the control input.

Figure 2.3. Closed loop response of a first order plus dead time process using P
controller with different controller gains (K).

Table 2.1
P PI PID
Ziegler-Nichols
KC KU/2 KU/2.2 KU/1.7
τI -- PU/1.2 PU/2
τD -- -- PU/8
Tyreus -Luyben
KC -- KU/3.2 KU/2.2
τI 2.2PU 2.2PU
τD -- -- PU/6.3

14
 It is appropriate to highlight that a controller is required to handle two types of
changes namely, a change in the output set-point and a change in the measured / unmeasured
disturbance into the process. The closed loop response for these is respectively referred to as
the servo and the regulator response. A disturbance into a process is also sometimes referred
to as a load change. Control systems in the process industry are typically designed for
effective load rejection. In contrast, set-point tracking is the primary objective in the design
of control systems for aerospace systems such as aeroplanes, rockets and missiles.
Figure 2.4 plots the regulator response for a unit step in the load variable with a P, PI
and PID controller tuned using the ZN and TL tuning rules for the first order plus dead time
process considered earlier. Notice that P only control results in an offset at the final steady
state. This offset is larger for TL tuning due to the lower controller gain. The PI and PID
regulator responses show no offset at the final steady state due to integral action. Also notice
that the aggressive ZN tuning results in a quicker but oscillatory return to the set-point for the
PI controller. These oscillations are suppressed by the D action in a PID controller. PID
control leads to a faster and smoother return to set-point due to the stabilizing effect of D
action. It is also highlighted that the TL tuning leads to a comparatively sluggish but non-
oscillatory response due to the more conservative tuning parameters. Large and sudden
changes in the control input are not desirable in the process industry to avoid hitting
operating constraints (e.g. flooding / weeping in sieve tray towers) during transients. Also,
the process equipment changes its dynamic characteristics due to equipment fouling, change
in process through-put, wear and tear over time etc so that the need for retuning a control
loop is mitigated using conservative controller settings. The TL settings thus represent a good
compromise between control performance and robustness.

Figure 2.4. Dynamics of manipulated and controlled variables using P, PI and PID
controllers with ZN and TL controller parameters for a unit step change in
load. (Regulatory response).

15
2.3. Process Identification

Obtaining the ultimate gain and period of a control loop by increasing the controller
gain causes the process to be driven towards instability. Considering the hazardous nature of
chemicals processed in any chemical plant, such a methodology for tuning loops must be
avoided. Alternative methods are needed that can be used for proper tuning. Two practical
methodologies namely, the process reaction curve and auto-tune variation are presented next.

2.3.1. Process Reaction Curve Fitting

The process reaction curve is the open-loop response of the output variable to a step
change in the manipulated variable which usually corresponds to a step change in a valve
position. Most of the transient responses can be well represented by a first order plus dead
time model. The model parameters are obtained as illustrated in Figure 2.5. The model
parameters can be obtained by two methods as illustrated in Figure 2.5. In both methods, the
ratio of the change in the controlled variable (output) from the initial to the final steady state
to the magnitude of the step change gives the process gain KP. For the controller, both input
and output are 4-20 mA signals corresponding to the sensor and final control element span. In
most commercial DCS systems, this range is represented as an equivalent 0-100% range. The
units of KP are then % change in controlled variable per % change in manipulated variable.

The two methods differ in the manner in which the dead time, θ, and the first order
time constant, τP, are obtained. In Method 1, a tangent at the inflection point in the process
reaction curve is drawn. Its intersection with the time axis gives the dead time θ. Its
intersection with the horizontal line Y = YSS, where Yss is the final steady state equals θ + τP,
from where τP is obtained. Equivalently, τP is obtained as

where S is the slope of the tangent drawn at the inflection point.

In Method 2, the time it takes for the response to reach 28.3% and 63.2% of the final steady
state are noted. Denote these two times with t28.3% and t63.2% respectively. Noting that for a
first order lag, 28.3% and 63.2% response completion occurs in τP/3 and τP time units
respectively, we have

θ + τP/3 = t28.3%

θ + τP = t63.2%

Subtracting the two equations to eliminate θ, we have

τP = 1.5(t63.2% - t28.3%)

and finally θ = 1.5 t28.3% - 0.5 t63.2%

16
The response of the fitted model using the two methods in shown in Figure 2.11. Method 2 is
clearly simpler and fits the actual process reaction curve better.

Fitting a First Order Plus Dead Time Model

2.5
Method 2
Method 1

2 θ + τP
Process Output / Input

1.5
0.632YSS Kp=YSS/U
YSS
1

0.283YSS
0.5 U

0 t28.3% t63.2%
θ

-0.5
0 10 20 30 40 50 60 70 80
Time

Figure 2.5. Fitting a first order plus dead time model to the process
reaction curve

With the fitted model, KU and PU can be obtained either by simulation or complex variable
analysis. The ZN or TL tunings can then be calculated as in Table 2.1.

17
2.3.2. Autotuning

Astrom and Hagglund (1984) proposed a powerful auto-tune variation (ATV) method
for obtaining the ultimate gain and ultimate period. The method consists of putting a relay at
the error signal that toggles the process input by ±h% on detecting a zero crossing. This is
schematically illustrated in Figure 2.6(a). The action of the relay causes the process input to
toggle around the steady state by ±h% for every zero crossing in the error signal
corresponding to the output crossing the set-point. Sustained oscillations result and the
system ends up in a limit cycle as depicted in Figure 2.6(b). The period of oscillations is the
ultimate period PU. The amplitude a of the output oscillations gives the ultimate gain KU as
4h
KU =
aπ
The ATV method has advantages over open loop step methods. The method automatically
finds the critical frequency (or period) of the process. Also, large deviations away from the
steady state are avoided as this is a closed loop test. Finally, the amplitude at the critical
frequency (ultimate period) is obtained so that the identification procedure is more accurate
than step / pulse tests.

+
Relay Process
input output

Figure 2.6(a). Block diagram of relay feedback approach

Figure 2.6(b). Relay feed back experiment a process with positive steady state gain

18
2.4. Controller Modes and Action

In all DCS systems, the controller can be in the indicator, manual, automatic or
cascade mode. In the indicator mode, the controller is off and the process variable (controlled
variable) is displayed. The control valve position cannot be adjusted by the operator. In the
manual mode, the controller is off. The process variable reading is displayed and the operator
can manually input the control valve position. Open loop step / pulse tests are performed in
the manual mode with the operator giving a step change to the control valve position. In the
automatic mode, the controller is on so that the control valve position is now set by the
controller. The operator inputs the set-point for the controlled variable. In the cascade mode,
the controller receives the set-point for the controlled variable from a master controller (and
not the operator).
Depending on the sign of the process gain, the controller action must be specified to
be “direct” or “reverse”. Usually a “direct” acting controller increases the controller output as
the controlled variable increases above the set-point. A reverse acting controller, on the other
hand, decreases the controller output as the controlled variable increases above set-point. For
a negative process gain, the controller is “direct” acting while for a positive process gain the
controller is “reverse” acting. The definition of “direct” or “reverse” action can vary from one
vendor to the other and it is always best to confirm the definition. Another consideration in
correctly specifying the controller action is whether the control valve fails open (air-to-close)
or fails closed (air-to-open). Process safety considerations dictate if a control valve fails open
or fails closed. For example the cooling water valve for removing heat from a reactor would
fail open while the steam valve into a reboiler would fail close. If the controller action for a
fail open valve is “direct”, the action would be “reverse” for a fail close valve in the same
control loop.
In control parlance, the controller gain is many-a-times reported as proportional band.
The proportional band is defined as
100
PB = %
KC
The higher the proportional band, the lower the controller gain and vice versa.

2.5. Rules of Thumb for Controller Tuning

Almost all control loops in the process industry are one of the following
Flow control loop
Pressure control loop
Level control loop
Temperature control loop
Product quality control loop
Some heuristics are discussed for tuning these loops that reflect common industrial practice.
Depending on the application, exceptions to these heuristics are always possible.

2.5.1. Flow Loops

Flow is usually controlled using a PI controller. The signal from the flow sensor is
noisy due to turbulent flow so that a large proportional band (about 150%) is used. A small
reset time (10-20s) is used for good set-point tracking.

19
2.5.2. Level Loops

Most liquid levels provide surge capacity for filtering out flow disturbances. For
example, the reflux drum in a distillation column allows for the reflux into the column to be
held constant even as the vapour condensation rate and distillate rate vary. If the drum is not
provided, the reflux into the column would fluctuate unnecessarily disturbing the column.
The reflux drum thus acts as a surge capacity. In order to filter out flow disturbances, the
level should be controlled loosely. The control objective is to maintain the liquid level within
acceptable limits. Accordingly, a P controller is used for level control. A proportional band of
50% is commonly used so that the valve fully closes / opens for a 25% change in the level
assuming the valve is initially 50% open. Note the use of PI controllers for level control of
surge capacities is not recommended as a change in the inlet (outlet) flow would require that
the outlet (inlet) flow increase above (decrease below) the inlet flow before becoming equal
to the inlet flow in order to bring the level back to its set-point (zero offset). The flow
disturbance thus gets magnified downstream (upstream). This magnification would only
worsen for a series of interconnected units defeating the very purpose of providing surge
capacity for attenuating flow disturbances. There are, of course, exceptions where tight level
control is desired. For example, the level in a CSTR should be controlled tightly to maintain
the residence time.

2.5.3. Pressure Loops

The dynamics of pressure in a can be very fast (flow like) or slow (level like)
depending on the process system. For example, the pressure dynamics are extremely fast for a
valve throttling the vapour outlet line from a tank. On the other hand, the dynamics are slow
for the cooling water flow adjusting the pressure in a condenser due to the heat transfer and
water flow lag. PI controllers are usually used for pressure loops with a small proportional
band (10-20%) and integral time (0.2-2 mins) for tight pressure control. Tight pressure
control is usually desired in most processing situations. For example, in distillation columns,
the pressure must be controlled tightly as large pressure deviations would require
compensation of the temperature controller set-points that ensure inferential product quality
control. Similarly, most gas phase reactors are designed for near maximum pressure operation
for maximum reaction rates so that large pressure deviations are not acceptable.

2.5.4. Temperature Loops

Temperature loops are moderately slow due to sensor lags and heat transfer lags. PI
and PID controllers are often used. In most processing situations, tight temperature control is
desired so that the proportional band is low (2-20%). The integral time is usually set to about
the same value as the process time constant. In situations where derivative action is used for
faster closed loop response, the derivative time constant is set to about one-fourth the process
time constant or less depending on the transmitter signal noise.

2.5.5. Quality Loops

Composition control loops are usually applied for maintaining the product quality. In
terms of relative importance, these loops are probably the most crucial for process
profitability. If the product quality shows large variability, the process must be operated at a
mean product quality that is significantly better than the quality specification to ensure the
production of on-spec or better quality product all the time. This results in a quality giveaway

20
adversely affecting the process profitability. The quality giveaway can be reduced by
ensuring tight product quality control. The concept of quality give-away is illustrated in
Figure 2.7.
Typical composition measurements involve large dead-times or lags. For example the
dead-time introduced by a gas-chromatograph can vary from a few minutes to an hour. Some
compositions may be measured once a shift or once a day through laborious analytical
measurements. Of all the measurements, analytical composition measurements are the most
expensive and unreliable. The product specifications increasingly require the measurement of
ppm / ppb levels of trace impurities so that a logarithmic scale is more appropriate in many
situations. Product quality measurements are typically used to make small / incremental
adjustments in the set-point of a loop. The frequency of the changes may vary from once a
day to once every hour etc. Whenever PID controllers are applicable, a large proportional
band is used (100-2000%). A large reset time (0.1 – 2 hrs) must be used due to the lag
introduced by the composition measurement as well as the usually slow process dynamics.

Figure 2.7. The concept of quality give-away

21
 Chapter 3. Advanced Control Structures
The feedback control loop, discussed at length, forms the backbone of control systems
applied in the process industry. Some typical feedback control loops are schematically
illustrated in Figure 3.1. Over the years, enhancements to the basic feedback control structure
that lead to significant improvement in control performance, have been developed. These
advanced control structures include ratio control, cascade control, feed-forward control, over-
ride control and valve positioning control and are briefly described in the following.

(a) (b)

FC FT

LT

LC

(c)
(d)
PT PC

TT TC

Figure. 3.1. Typical feed back control schemes commonly employed in distillation
columns. (a) Feed flow control, (b) Level control in reboiler drum using
bottoms flow, (c) Tray temperature control using reboiler duty and (d) Column
pressure control using condenser duty.

3.1. Ratio Control

Ratio control, as the name suggests, is used for maintaining the ratio between two
streams. The independent stream is referred to as the wild stream. The ratio controller adjusts
the flow of the other stream to keep it in ratio to the wild stream. The implementation of ratio
control is illustrated in Figure 3.2. The wild stream flow measurement is multiplied by the
ratio set-point to obtain the flow set-point for the manipulated stream. The calculated flow
set-point is input to the flow controller on the manipulated stream. Ratio control is
implemented as a feed-forward strategy (to be discussed later) where two flows are increased

22
in tandem so that the change in the wild stream is compensated for before it affects the
process output. For example, if the feed flow rate into a distillation column increases by 10%,
the reboiler duty necessary to maintain the same separation should also increase by about
10%. It therefore makes sense to ratio the reboiler duty to the fresh feed rate so that the
necessary change in the reboiler duty is implemented apriori. This leads to tighter product
purity control with the change in the feed rate causing only small deviations in the product
purity.

Wild stream

FT

Multiplier X Constant

Flow set point

Flow controller FC

FT

Manipulated stream

Figure. 3.2. Implementation of ratio control.

3.2. Cascade Control

Cascade control is arguably one of the most useful concepts in chemical process
control. The cascade control scheme consists of two control loops, namely the master loop
and the slave loop, with the master loop setting the set-point for the slave loop. The concept
is best illustrated by an example. Consider a jacketed CSTR where cooling water is
recirculated in the jacket to remove the exothermic reaction heat. The typical feedback
reactor temperature control scheme and the cascade reactor temperature control scheme is
shown in Figure 3.3. In the feedback arrangement, the reactor temperature controller directly
adjusts the cooling water valve to maintain the reactor temperature at set-point. In the cascade
arrangement, a slave loop is introduced that controls the jacket temperature by manipulating
the cooling water valve. The master reactor temperature loop adjusts the jacket temperature
set-point.
At first glance, the advantage of cascade arrangement over simple feedback control is
not very obvious. To appreciate the same, consider an increase in the coolant temperature as
an input disturbance. In the simple feedback scheme, the reactor temperature must rise before
the controller opens the cooling water valve to bring the reactor temperature back to set-
point. In the cascade control scheme, the jacket temperature controller senses the increase in
the cooling water temperature and adjusts the cooling water valve to maintain the jacket
temperature. The reactor temperature would thus show comparatively much smaller /
negligible deviations from set-point. The slave controller acts to remove local disturbances
into the process and prevents its effect on the primary controlled variable. Another subtle

23
advantage is that the slave controller compensates for the non-linearity in the slave loop so
that the master controller ‘sees’ a more linear system. In the current example, the non-linear
characteristics of the cooling water valve are compensated for by the slave controller. Since
the slave loop has much faster dynamics than the master loop (else the cascade arrangement
is infeasible), the master loop does not have to compensate for the valve non-linearity. It
therefore sees a less non-linear system compared to simple feedback control resulting in
improved control performance. The improvement is however at the expense of installing,
tuning and maintaining an additional slave controller.

(a)

TT

LC

TC

(b)
TC
TT

LC

TC TT

Figure 3.3. Temperature control of an exothermic CSTR. (a) the typical feedback reactor
temperature control scheme and (b) the cascade reactor temperature control scheme.

To tune a cascade control structure, the slave loop is first tuned with the master loop
in manual. P only controllers with a small proportional band (large controller gain) are
commonly used in the slave loop for a fast response to a set-point change from the master
controller. Integral action is usually not applied in the slave loop as an offset in the secondary

24
measurement is acceptable. The tuned slave loop is then put on automatic and the master loop
is tuned. Note that for the cascade control system to be stable, the dynamics of the slave loop
should be much faster than the master loop allowing the slave loop to keep-up with the set-
point changes received from the master loop. A typical rule of thumb is that the time constant
for the master loop should be more than thrice that of the slave loop.
Cascade control loops are quite common in the process industry. Some common
configurations are shown in Figure 3.4. The interpretation of these configurations is left as an
exercise to the reader.

TC
SP

xD FC

TC CC
SP

Figure 3.4. Some typical cascade arrangements

3.3. Feed-forward Control

The concept of feed-forward control has already been alluded to earlier. If a measured
disturbance enters a process, the control input can be adjusted to compensate for effect of the
disturbance on the output. Perfect compensation would cause the controlled output to show
no deviations from its set-point even as a disturbance has entered the process. This apriori
compensation to mitigate the transient effect of a measured disturbance on the controlled
output is referred to as feed-forward control. A very simple example of feed-forward control
is driving a car. Adjusting the hot and cold water knobs for the right temperature water from
the shower is an example of feedback control. As discussed previously, ratio control
compensates for disturbances in a feed-forward manner.
The design of a feed-forward compensator is illustrated using block diagrams in
Figure 3.5. Gd represents the disturbance to output transfer function while Gp represents the
control input to output transfer function. The control input u must be varied such that
Gp.u + Gd.d = 0
The control input is adjusted by the feed-forward compensator with the transfer function Gff
so that
u = Gff.d.
Substituting into the previous equation and solving for Gff gives the feed-forward
compensator design as
Gff = -Gd/Gp
Assuming that Gd and Gp are first order plus dead time transfer function, the feed-forward
compensator is then a lead-lag plus dead time transfer function. Modern DCS allow lead-lag
plus dead time blocks to be configured into the control system.
For a better appreciation of the improvement in control performance using feed-
forward compensation, consider a very simple example where

25
 Gd = 1/(s+1)
and Gp = 1/(5s+1)

Then Gff = -(5s+1)/(s+1)

Figure 3.6. plots the simulated transient output response for a unit step change in the
measured disturbance with and without feed-forward compensation.
Since there is no plant-model mismatch, perfect feed-forward compensation is observed with
the output showing no deviations from set-point. In a real-life scenario, the presence of a
plant-model mismatch may cause small transient deviations. The feed-back controller
compensates for these small deviations resulting in an overall tighter closed loop response.
(a) d

Gd
Disturbance

u(s) y(s)
Input GP Output
Process
(b)
d

Gff = - Gd / GP Gd
Disturbance

y(s)
Input GP
Output

Figure 3.5. Design of feed forward compensator. (a) Process and (b)
process with feed forward compensator.

Figure.3.6. Deviation in the output with and without feed forward action
26
3.4. Override Control

Over-ride control is employed to ensure that an unsafe condition does not arise during
process operation. As the name suggests, an over-ride controller over-rides the output of
another controller as an unsafe condition develops and acts to move the process away from
the unsafe condition. This is an example of multivariable control where the same manipulated
variable can be adjusted at any time by one of many controlled variables. An example best
illustrates the concept of over-ride or selective control. Consider the bottom section of a
distillation column. The bottom sump level is controlled by the bottoms flow rate. During
normal operation, the steam rate into the reboiler is manipulated to control a tray temperature.
During severe transients, a situation may arise where the bottoms level is low and continues
to fall even as the bottoms flow rate is zero. An unsafe situation can arise with the reboiler
tubes getting exposed to vapour and fouling. Also, the bottoms pump may lose suction as the
reboiler dries up. A sensible operator would put the temperature loop on manual and cut back
on the steam rate to ensure the reboiler tubes remain submerged. In effect, the temperature
controller output, the signal to the steam valve, gets over-ridden to maintain the liquid level.
The over-ride controller automates this action as shown in Figure 3.7. The base level signal is
input to a multiplier. A multiplier value of 5 is used so that if the level is above 20%, the
multiplier output is above 100%. As the level decreases below 20%, the multiplier output
decreases below 100%. If the level continues to decrease, the multiplier output would
eventually decrease below the temperature controller output. The low select would then pass
on the multiplier signal to the steam valve over-riding the temperature controller. The steam
rate would thus decrease. Once the level begins to rise, the multiplier output would increase
above the temperature controller output so that the low select would pass the manipulation of
the steam valve back to the temperature controller. In addition to the level over-ride
controller, the low select may also receive signals from a pressure over-ride controller or a

TC LS

TT

Steam

LT

Bottoms Low base

level
override
controller
LC

Fig. 3.7. Override control scheme

27
temperature over-ride controller to reduce the steam flow rate. Pressure over-ride would be
needed if the column pressure goes too high. Similarly temperature over-ride may be
necessary if the base temperature goes too high.
In temperature or pressure over-rides, a PI controller is needed unlike the P only
controller for a level over-ride. This is because a pressure / temperature over-ride is needed
only for a very small range of the total transmitter span. A very large proportional gain would
then be necessary which can destabilize the closed loop system. Therefore a PI controller
with lower gain and fast reset action is used to achieve the tightest control possible.

3.5. Valve Positioning (Optimizing) Control

Valve positioning control was originally proposed by Shinskey as an effective way of

minimizing the energy consumption in distillation columns. The pressure in a distillation
column is set by the condenser cooling duty. For a given separation, as the column pressure
increases, more stages are needed as the x-y VLE plot moves towards the 45 degree line as
shown in Figure 3.8. Translated to process operation, the same separation can be achieved at
lower reboil as the column operating pressure is reduced. To minimize energy consumption,
the column should be operated at lowest possible pressure corresponding to the maximum
condenser duty. This can be accomplished by the valve positioning control scheme as
illustrated Figure 3.9. The column pressure is typically controlled by adjusting the condenser
cooling water valve. The VPC controller takes in the pressure controller output signal and
adjusts the pressure set-point. If the valve is not nearly open, the controller reduces the
column pressure set-point so that the pressure controller increases the cooling duty to reduce
the column pressure. The VPC controller thus ensures that any underutilized cooling capacity
is exploited to reduce the column operating pressure. The column pressure thus floats with
the condenser duty being near maximum. The VPC controller is tuned to be slow with the
fast pressure controller rejecting any pressure disturbances.

0
P
P : Pressure
x : Liq. mole fraction
y : Vap. mole fraction

1
0 x 1
Figure 3.8. Typical x-y diagram with varying Pressures

28
 SP = 95%

SP

Figure 3.9.Valve positioning control

Another simple VPC application is shown in Figure 3.10. Let us say a high capacity variable
speed pump is providing feed to N parallel trains of processes. We would like to minimize
the pump electricity consumption while ensuring the desired flow setpoints for each of the
parallel trains is achieved. The electricity consumption gets minimized by running the pump
at as low an rpm as possible. This gets achieved by ensuring that the most open process feed
valve is nearly fully open. The high select passes the position of the most open valve. A valve
position below the nearly fully open VPC setpoint (say 80%) indicates unnecessary valve
throttling. The VPC then reduces the pump rpm. In response, the flow controllers would open
the valves to maintain the flow. The VPC reduces the pump rpm till the most open valve
position reaches the VPC setpoint (80%) ensuring the pump operates at as low an rpm as
possible while maintaining the desired flow to each of the parallel trains.

SP =80% VPC F1SP FC

HS

SC Train 1
F2SP FC

Train 2

FNSP FC

Train N

Figure. 3.10. VPC for minimizing variable speed pump electricity

consumption

29
 Chapter 4. Multivariable Systems

Single input single output (SISO) systems have been treated till now. Most practical
control system design problems are multivariable in nature with multiple inputs multiple
outputs (MIMO). A 2 X 2 multivariable system is shown in Figure 4.1. There are two inputs,
u1 and u2 and two outputs y1 and y2. In the most general case, a step change in an input causes
a transient response in both the outputs. The input output relationship may be compactly
represented in matrix notation as
 y1 ( s )  G11 ( s ) G12 ( s )   u1 ( s ) 
 y ( s) = G ( s ) G ( s ) u ( s )
 2   21 22  2 
and the corresponding block diagram is shown in Figure 4.1.

+
G11(s)
y1(s)
u1(s) +

G21(s)

G12(s)

+
G22(s)
u2(s) + y2(s)

Figure 4.1. A block diagram of a 2 X 2 multi variable system

In general, Gij denotes the transfer function between the jth input and the ith output.
The non-diagonal terms with i ≠ j are the interaction terms. The simplest way of controlling a
multivariable process is to control each of the outputs by manipulating an input using a PID
controller. This is referred to as multivariable decentralized control and is illustrated in Figure
4.2. for the example 2x2 system. Controller 1 manipulates u1 to maintain y1 and controller 2
adjusts u2 to maintain y2.
In the design of a multivariable decentralized control system, choice exists as to
which manipulated variable is used to control an output. For the 2x2 example, there are a
total of two control structures with y1 being controlled by u1 or u2. The number of such
possibilities grows exponentially as the number of inputs / outputs increase. In the most
general sense, the design of a plant-wide decentralized control system for a complex chemical
process is a multivariable problem of high order. The high order problem is naturally broken
down into smaller process unit specific controller design problems and controller design for
managing plant-wide issues such as inventory balancing. A high order unit specific controller
design problem can also be further broken down into a smaller subset of fast loops and slow
loops based on the process dynamics. An example is the simplification of the 5x5 controller
design problem for a simple distillation column into a 2x2 problem. In a distillation column,
the pressure, reflux drum and bottom levels and two temperatures (or compositions) may be
controlled. Since the tray temperature dynamics are significantly slower than the pressure /

30
level dynamics, SISO controllers are applied for the latter reducing the 5x5 problem into a
2x2 design problem for the two temperature controllers. Any complex high order control
system design problem can thus be simplified into subsets of simple SISO, 2x2 or in the
worst case 3x3 decentralized control system design problems. A systematic unit specific and
plant-wide control system design methodology for complete chemical plants will be
developed in the subsequent chapters.

y1

y1, SP +
GC1(s)
G11(s)
+ u1(s) +
y1(s)

G21(s)

G12(s)

y2, SP +
+ Gc2(s)
G22(s)
u2(s) + y2(s)
y2

Figure 4.2. Block diagram of a multivariable decentralized control for a 2X2 system

4.1. Interaction Metrics

The selection of the input-output pairing in a decentralized control system is usually

made based on engineering considerations which shall be covered in greater detail in
subsequent chapters. The individual controllers in a decentralized control system may need to
be detuned in order to maintain process stability. This is because the interaction between the
loops during closed loop operation can lead to instability. The magnitude of interaction
depends on the aggressiveness of the individual controller tunings employed. Detuning or
less aggressive tuning mitigates the interaction to ensure closed loop stability. The
Niederlinski Index and Relative Gain Array are two commonly used quantitative measures of
interaction between control loops. Both are based on the open-loop steady state gain matrix
KP, where
y = KP u

4.1.1. Niederlinski Index

The Niederlinski Index for a control structure where the ith input is used to control the
th
i output is then defined as

31
 KP
NI =
∏K
i
ii

The NI for any control structure can thus be obtained through appropriate relabeling of the
outputs and inputs so that the ith input controls the ith output. If the Niederlinski Index is
negative, the closed loop system is guaranteed to be integral closed loop unstable. If the NI is
positive, the closed loop system may or may not be stable. In other words, the criteria NI>0 is
a necessary but not sufficient condition for closed loop stability. Input-output pairings with
small positive or large positive (>>1) NI values indicate ill-conditioning problems and should
be avoided. Control structures with NI close to 1 indicate favourable interaction. For
example, an NI value of 1 for a 2X2 system indicates that either K12 or K21 or both are zero
implying one-way or no steady state interaction between the loops. The primary use
Niederlinski Index is for rejecting unworkable control structures.

4.1.2. Relative Gain Array

The relative gain is another popular metric that measures the interaction of a control
loop with other loops as the ratio of the steady state process gain the controller sees with all
other loops off to the process gain with all other loops on (all other outputs at their set-
points). Mathematically, if the ith output is controlled by the jth input, its relative gain is
defined as
 ∂y i 
 
 ∂u 
 j  uk = cons tan t ,k ≠ j
λij =
 ∂y i 
 
 ∂u 
 j  yk = cons tan t ,k ≠i
If the relative gain is negative, the ith output should not be paired with the jth input as the
process gain sign would change depending on whether the other loops are on automatic or
manual mode. Input-output pairings with relative gain close to 1 may be preferred as the
process gain the controller sees is independent of the state of the other loops. The relative
gain array is obtained as i and j are varied for respectively all outputs and inputs.
The relative gain array is an effective tool for input-output pairing when the primary
control objective is set-point tracking. For set-point tracking, lower interaction between the
loops increases the degree of independence of the different control loops so that each can be
separately tuned for tight set-point tracking. Interaction is thus undesirable for set-point
tracking. For load disturbance rejection, interaction is not necessarily undesirable and may
actually favour disturbance rejection. This was demonstrated in an early article by
Niederlinski (1971). Since the primary objective in chemical process control is load rejection,
the application of RGA for control structure selection makes little sense. Candidate control
structures should be proposed based on engineering considerations and unworkable structures
further eliminated using the Niederlinski Index. The same arguments can be applied to
recommend the use of dynamic decouplers only when the primary control objective is set-
point tracking. Dynamic decoupling is not covered here as load rejection is the primary
control objective in chemical process control systems.

32
4.2. Multivariable Decentralized Control

Consider the 2x2 multivariable open loop system in Figure 4.1. We would like to hold
both the outputs at their respective setpoints. The simplest way to do it is to implement
individual PI controllers for y1 and y2. Without loss of generality, let us assume that y1 is
paired with u1 and y2 is paired with u2. The multivariable control system is shown in Figure
4.2. Notice that even as u1 and u2 affect both y1 and y2 through the interaction transfer
functions G12 and G21, the adjustment made to u1 is based purely on e1 and the adjustment
made to u2 is based purely on e2. In other words, the y1 controller moves are based purely on
y1 and does not consider the effect of the control moves made by the y2 controller. Similarly,
the y2 controller moves are based purely on y2 and does not consider the effect of control
moves made by the y1 controller. Thus even as the actual system is multivariable, the
individual controllers do not take the interaction into consideration. This is referred to as
decentralized control.
For the decentralized control system, notice that the interaction terms introduce an
additional feedback path as shown in blue in Figure 4.3. This additional feedback tends to
further destabilize the closed multivariable control system. If each controller is tuned
individually with the other controller on manual (other loop is open) and the Zeigler Nichols
tunings applied, then when both the loops are closed, the system response is likely to be
highly oscillatory and may even be unstable due to the additional feedback path. In the
individual tuning of the controllers, since the other loop is open, this additional feedback path
is inactive and therefore not accounted for in the determination of the tuning parameters.
Clearly the individual ZN tuning parameters need to be detuned due to the additional
feedback path to ensure the overall closed loop response is sufficiently away from instability.

4.2.1. Detuning Multivariable Decentralized Controllers

The obvious next question is that how does one tune a decentralized multivariable
controller. Typically, in practical settings, tight control of one of the outputs is much more
important than the other. A sequential tuning procedure can then be applied, where the more
important output controller is tuned individually so that we get the tightest possible controller
tuning. The less important output controller is then tuned with the other loop on automatic.
Since the other loop is on, the additional feedback path is active and the necessary detuning
due to the same gets accounted for in the tuning parameters of this less important loop. This
sequential tuning procedure thus gives the tightest possible control of the more important
output at the expense of a highly detuned controller for the less important output. The
sequential procedure can be easily extended to more than 2 outputs when the prioritization of
the controlled outputs is clear.
There are however situations where the need for tight control of each of the outputs is
comparable. The detuning due to multivariable interaction then needs to be taken in all the
loops. How does one systematically go about the detuning. For the 2x2 multivariable system,
we have for the open loop system

or more simply
y = GP u

where GP is the open loop process transfer function matrix. For a decentralized controller, we
have

33
or in matrix notation
u = GC (ySP – y)
where the controller matrix, GC, is diagonal for decentralized control. Combining the above
two matrix equations, we get
y = GP GC (ySP – y)
or (I + GP GC) y = GP GC ySP
or y = (I + GP GC)-1 GP GC ySP
This is the multivariable closed loop servo response equation and its analogy with SISO
systems is self evident. Each element of the (I + GP GC)-1 matrix would have det(I + GP GC)
as its denominator. The closed loop multivariable characteristic equation is then
det(I + GP GC) = 0
Similar to SISO systems, if any of the roots of the multivariable characteristic equation is in
the right half plane, the closed loop multivariable system is unstable.
To systematically detune the controllers, an empirical analogy with the Nyquist
stability criterion for SISO systems is used. For a SISO system, the closed loop servo
response equation is
y = [GP GC/(1 + GP GC)] ySP
where GP is the open loop transfer function and GC is the controller transfer function. The
Nyquist stability criterion then guarantees stability for the closed loops system if the polar
plot of the open loop transfer function between ySP and y, ie GPGC, does not encircle (-1, 0).
Gain margin and phase margin are criteria that are commonly used to quantify the distance
from (-1, 0) at a particular frequency. To ensure that the distance from (-1, 0) is sufficient at
all frequencies, the 2 dB closed loop maximum log modulus criterion is often used, where the
closed loop log modulus is defined as
LCL(ω) = 20 log|GPGC / (1+GPGC)|s=jω
LCL is calculated by putting s = jω in the transfer functions, GP and GC, and is therefore a
function of ω. The SISO PI tuning parameters (KC and τI) are chosen such that the maximum
closed loop log modulus (with respect to ω) is 2dB. This ensures that the closed loop servo
response is fast and not-too-oscillatory.
To develop a closed loop maximum log modulus criterion for multivariable systems,
we note that the SISO closed loop characteristic equation is
1 + GPGC = 0
and the transfer function whose polar plot is used to see encirclements of (-1,0) is then
-1 + (1+GPGC)
ie -1 + closed loop characteristic equation
For a multivariable system, we then define by analogy
W = -1 + det(I + GP GC)
where W is -1 + closed loop characteristic equation. The multivariable closed loop log
modulus (LMVCL) is then defined as
LMVCL = 20 log|W/(1+W)|.
The tuning parameters for the individual controllers should be chosen such that
LMVCLMAX = 2 NC
where NC is the number of loops.
A simple algorithm for systematic detuning of the individual controller for the 2x2
decentralized control system is then:
1. Obtain individual ZN tuning parameters, (KC1ZN, τI1ZN) and (KC2ZN, τI2ZN), for each
loop.
2. Detune the individual tuning parameters by a factor f (f > 1) to get the revised tuning
parameters as (KC1ZN/f, f.τI1ZN) and (KC2ZN/f, f.τI2ZN)
3. Adjust f such that LMVCLMAX = 4 dB.

34
The above procedure can be easily extended to an NxN (N > 2) decentralized control system.

As a parting thought, we re-emphasize that in chemical processes, the dominant time

constants of different loops can differ by up to two orders of magnitudes. Thus for example,
the residence time of a surge drum may be ~5 minutes while it may take 2-5 hrs for transients
caused by a change in its setpoint to reach back after passing through the different
downstream units, the material recycle and the upstream units. Similarly, on a distillation
column, while the column pressure time constant with respect to condenser duty is ~1 min
and the reflux drum / bottom sump level residence times are ~ 5 mins, the tray temperature
response times to changes in reflux / boilup rates are much slower (~15-20 mins). Thus even
as the dual-ended distillation column control problem is 5x5 (2 levels, 1 pressure and 2
temperatures), the separation in time constants allows the level and pressure controllers to be
tuned first followed by the two temperature controllers. The 5x5 problem thus reduces to a
2x2 problem due to the separation in time constants. In industrial practice, most high order
multivariable problems reduce to 2x2 or at most 3x3 problems, which are mathematically
tractable.

y1

y1,sp +
Gc1(s)
G11(s)
+ u1(s) +
y1(s)

G21(s)

G12(s)

y2,sp +
+ Gc2(s)
G22(s)
u2(s) + y2(s)
y2

Figure 4.3. Additional feedback path due to multivariable interaction

35
Illustrative Example:
Consider a 2x2 openloop multivariable system

(a) Calculate its RGA. Based on the RGA, what input-output pairing would you recommend.
(b) Calculate the Niederlinski Index for the recommended pairing. What can you say about
closed loop integral stability of the recommended pairing.
(c) Calculate the Niederlinki Index for the other alternative pairing (the one that is not
recommended). What can you say about the closed loop integral stability of this other
pairing.
(d) For the recommended pairing, design a feedforward dynamic decoupler showing its
complete block diagram and also the physically realizable feedforward compensator
transfer functions.

Solution:
(a) The steady state input-output relationship is

so that the steady state gain matrix is

Inverting the matrix, we get

The RGA is then obtained as

RGA = K.*(K-1)T
where the ‘.*’ operator denotes element-by-element multiplication. Performing the necessary
operations, we get

Notice that the row/column sum of the RGA is 1. This is a property of the RGA (can you
prove it?).
Rejecting the IO pairings corresponding to the negative RGA elements, the recommended
pairing based on the RGA is y1-u2 and y2-u1.

(b) The steady state IO relation for the recommended pairing is

The Niederlinski Index is then

Since NI > 0 for the recommended pairing, the multivariable decentralized control system
may be integrally stable.

(c) The other possible pairing is y1-u1 and y2-u2. For this pairing, the IO relation is

36
The Niederlinski Index is then

Since the NI for this pairing is < 0, the multivariable decentralized control system is
guaranteed to be integrally unstable. This pairing should therefore not be implemented.

(d) If we look at the open loop 2x2 system with the recommended pairing (y1-u2 and y2-u1), a
change in u2 affects both y1 (its controlled variable, CV) and y2 (other CV). Similarly, a
change in u1 affects both y2 (its CV) and y1 (other CV). When both the control loops are on,
the adjustment made by a loop ends up disturbing the other loop. A dynamic decoupler uses
feedforward compensation ideas to make appropriate adjustments in the “other” process input
so that a change in a process input only affects its CV and not the other CV. The dynamic
decoupler block diagram for the recommended pairing is shown in Figure 4.4. We are
looking for the feedforward compensator GIff (GIIff) so that a change in u2* (u1*) only affects
its CV, y1 (y2) with no effect on the other CV y2 (y1).

u2 + u2 + y1
* +
+

GIff

GIIff

+ +
u1 + + y2
* u1
Decoupler
Process
Figure 4.4. 2x2 process example with dynamic decoupler
From the block diagram, the ideal compensator GIff would be such that
y2 = G22u2* + G21GIffu2* = 0
so that GIff = -G22/G21
Similarly, we have GIIff = -G11/G12
Putting in the appropriate transfer functions, we get

37
The feedforward compensators consist of a gain, a lead-lag and a deadtime. In some cases, it
is possible that we get an exponential term of form e+Ds (D > 0) implying a negative dead-
time. This means that a change in the causal variable leads to a change in the effected
variable in the past, which is impossible. The term e+Ds is then physically unrealizable and
dropped from the compensator.

38
 MODULE II

CONTROL OF COMMON UNIT OPERATIONS

Having covered the essential aspects of control theory, in this module we consider control
systems as applied to common unit operations in the process industry. We thus treat simple
and complex distillation configurations (including heat integrated sequences), reactors, heat
exchangers and miscellaneous systems such as furnaces, compressors, refrigeration cycles
and boiler houses. Several of the examples shown here can be found in “Plantwide Process
Control” by Luyben, Tyreus and Luyben (McGraw Hill, 1998) and “Process Control
Systems” by Shinskey (McGraw Hill, 1996). We have attempted to present these examples
afresh in the hope that readers readily assimilate the concepts.

39
 Chapter 5. Control of Distillation Systems
5.1. Distillation Basics

Distillation is unarguably the most preferred unit operation used for separating
mixtures. In the design of chemical processes, other separation techniques are considered
only if distillation is found to be economically unviable. It is thus not surprising that the final
product stream from a plant is typically a product steam from a distillation column. This
Chapter provides guidelines for designing effective control systems for distillation columns.

5.1.1. The Simple Distillation Column

A proper understanding of the basic physics of a distillation column (or any other
process for that matter) is a pre-requisite for designing an effective control system. Figure 5.1
shows the schematic of a simple distillation column along with the control valves. It consists
of a tray section, a condenser, a reflux drum and a reboiler. The feed mixture is fed on a feed
tray. The trays above the feed tray constitute the rectifying / enriching section and those
below constitute the stripping section. The overhead distillate and the bottoms are the two
product streams from a simple distillation column. Steam is typically used to provide vapour
reboil into the stripping section. The liquid reflux into the enriching section is provided by the
condenser. Cooling water is commonly used as the coolant in the condenser. The condenser
may be a total condenser, where all the vapour is condensed, or a partial condenser where
only a part of the vapour is condensed. The overhead distillate is a liquid stream for a total
condenser. A partial condenser column may be operated at total reflux where all the liquid is
refluxed back into the column and the distillate stream is a vapour stream. Alternatively (and
more commonly) both a vapour and a liquid distillate stream are drawn. The reflux drum
provides surge capacity to adjust the reflux and distillate rate during transients. The bottom
sump provides the surge capacity for adjusting the bottoms and steam rate.
The vapour generated when a volatile liquid feed mixture is boiled is richer in the
more volatile component. The remaining liquid is then richer in the heavier components.
Chemical engineers refer to this as flashing a mixture. If the flashed vapour is condensed and
partially vaporized again, the vapour from the second flash would be further enriched in the
volatiles (light boilers). Similarly, if the liquid from the first flash is further vaporized, the
heavies composition of the liquid from the second flash would increase. Theoretically
speaking, a sufficiently large number of flash operations on the vapour can result in a final
vapour stream that is almost 100% pure lightest component. Similarly a series of flash
operations on the liquid can result in a final liquid product that is 100% heaviest component.
The array of trays in a distillation column accomplishes this series of flash operations. The
temperature difference between the liquid and vapour streams entering a tray causes
condensation / vaporization so that as one moves up the column, the composition of the
lightest component increases monotonically. Alternatively, as one moves down the column,
the composition of the heaviest component keeps on increasing. Since heavier components
boil at higher temperatures, the tray temperature increases as one moves down the column
with the condenser being the coolest and the reboiler being the hottest. The reboiler and the
condenser are the source of vaporization and condensation respectively for the series of
vaporization / condensation.

40
 Condenser

Reflux

Distillate
Feed

Reboiler Steam

Bottoms
Figure 5.1 Schematic of a simple distillation column along with the control valves.

5.1.2. Splits in a Simple Distillation Column

Consider a five component equimolar ABCDE mixture feed into a simple distillation
column. The components are in decreasing order of volatility so that A is the lightest and E is
the heaviest. The feed rate is 100 kmol/h. The steady state distillate to bottoms product split is
primarily determined by the choice of the distillate (or bottoms) rate. Assuming a sufficiently
large number of trays, adequate reboil and reflux, for a distillate rate of 40 kmol / hr, which is
equal to the component A and component B flow rate in the feed, essentially all of the A and
B would leave up the top so that the distillate would contain traces C, D and E impurities in
decreasing order of composition. The bottoms would be a CDE mixture with traces of B and
A, in decreasing order of composition. The column thus accomplishes a split between
components B and C with the liquid preventing C from escaping up the top and the vapour
reboil preventing B from escaping down the bottoms. Components B and C, are referred to as
the light key (LK) and heavy key (HK) respectively. The LK is the dominant impurity in the
bottoms stream and the HK is the dominant impurity in the distillate stream. The component
split is referred to as an AB/CDE split. The component that is the next lighter component
than the LK is called the lighter than light key (LLK). The heavier than heavy key (HHK) can

41
be defined in a complementary manner. Components A and E are respectively the lightest
and heaviest and therefore referred to as the lightest key and the heaviest key.

For the ABCDE mixture, there are four possible splits – A/BCDE, AB/CDE,
ABC/DE and ABCD/E. The first one, where the light key is also the lightest key is referred to
as the direct split. The last one, where the heavy key is also the heaviest key is referred to as
an indirect split. The remaining splits where the key components are intermediate boilers are
referred to as intermediate splits. It is helpful to categorize the column split into these basic
types.

5.2. Basic Control Structures

A simple distillation column with a total condenser has a total of six valves as in
Figure 5.1. Of these six valves, the feed valve is usually set by an upstream unit in the
process. Also two valves must be used to control the reflux drum level and the reboiler level
as liquid levels are non-self regulating. Another valve must be used to regulate the column
pressure which represents the vapour inventory in the column. Typically, the cooling duty
valve in the condenser is used for pressure control. After implementing the three inventory
loops, the position of the remaining two control valves can be set by an operator or a
controller to regulate the separation. This gives a operation degree of freedom of two for a
simple distillation column. The operation degree of freedom is more for complex column
configurations that are considered later.
Four control structure types result for a distillation column corresponding to the
choice of valve used for reflux drum and reboiler level control. These are the LQ, DQ, LB
and DB structures and are illustrated in Figure 5.2. The nomenclature corresponds to the two
control degrees of freedom (valves) that remain to regulate the separation. The LQ control
structure corresponds to the distillate (D) controlling the reflux drum level and the bottoms
(B) controlling the reboiler level. This leaves the reflux (L) and reboiler duty (Q) as the two
valves for regulating the separation achieved, hence the label LQ. In the DQ structure, the
condenser level is controlled using the reflux while in the LB structure, the bottoms level is
controlled using the reboiler duty. Lastly in the DB control structure, the reboiler duty and
reflux are used for controlling the reboiler and condenser levels respectively.

5.2.1. The Energy Balance (LQ) Structure

The LQ control structure is the most natural control structure for a simple distillation
column. This is because the separation in a distillation column occurs due to successive
condensation and vaporization of the counter-current vapour and liquid streams flowing
through the column. Adjusting the cold reflux, the source of condensation, and the reboiler
duty, the source of vaporization, is then a natural choice for regulating the separation
achieved in the column. The LQ control structure is thus the most commonly applied
distillation control structure. It is also sometimes referred to as an energy balance structure as
changing L (cold reflux) or Q alters the energy balance across the column to affect the
distillate to bottoms product split.

42
 (a) LQ PC (b) DQ PC

Condeser Condeser

L
FC
LC D

LC FC
FC Reflux FC Reflux

Distillate Distillate
Feed Feed

Q FC Q FC

LC Reboiler steam LC Reboiler steam

Bottoms Bottoms

(c) LB PC
(d) DB PC
Condeser
Condeser
FC
L D
LC
LC
FC Reflux FC
FC Reflux

Distillate
Feed Distillate
Feed

Reboiler steam
Reboiler steam
LC FC
B LC FC
B
Bottoms
Bottoms

Figure 5.2. Schematics of LQ, DQ, LB and DB control structures

43
5.2.2. Material Balance Structures

The other control structures are referred to as material balance structures as the
product split is directly adjusted by changing the distillate or bottoms stream flow rate. The
material balance structures are applied when a level loop for the LQ structure would be
ineffective due to a very small product stream (D or B) flow rate. The DQ structure is thus
appropriate for columns with very large reflux ratio (L/D > 4). The distillate stream flow is
then a fraction of the reflux stream so that the reflux drum level cannot be maintained using
the distillate. The level must then be controlled using the reflux. The LB structure is
appropriate for columns with a small bottoms flow rate compared to the boil-up. The bottoms
stream is then not appropriate for level control and the reboiler duty must be used instead.
The DB control structure is used very rarely as both D and B cannot be set independently due
to the steady state overall material balance constraint. In dynamics however, the control
structure may be used when the reflux and reboil are much larger than the distillate and
bottoms respectively. This occurs in super fractionators which will be discussed later in this
Chapter.

5.2.3. Other Control Structure Variants

Other variants of the basic control structure types include the L/D-Q, L/D-B and D-
Q/B. In the first two structures the reflux ratio is adjusted for regulating the separation. In the
last structure the reboil ratio is adjusted. These control structures are illustrated in Figure 5.3.
Note that when the reflux is adjusted in ratio with the distillate, the distillate stream can be
used to control the reflux drum level even as it may be a trickle compared to the reflux rate.
Similarly, when the reboil rate is adjusted in ratio with the bottoms, a small bottoms stream
can provide effective level control.
Maintaining the reflux ratio is quite common in distillation control as it provides
feedforward compensation in the reflux for a change in the distillate rate. Such a feedforward
compensation can significantly improve quality control as the column dynamics are slow
with respect to a change in the reflux rate due to the slow liquid hydraulics with every tray
having a time constant of 15-30 s. Pure feedback adjustment of the reflux can thus result in
large purity deviations. Maintaining the reboil ratio is not very popular. This is because all
the tray compositions / tray temperatures respond almost immediately to a change in the
reboil due to the fast vapour dynamics. Adjustment of the reboiler duty in a feedback
arrangement is thus usually sufficient for effective regulation.

5.3. Temperature Based Inferential Control

The distillation column performs a separation between the light key and the heavy key
so that heavy key and light key impurity levels respectively in the distillate and bottoms are
below design specifications. The primary control objective then is to ensure these impurity
levels remain below specifications for load changes. A change in the flow rate and
composition of the feed into the column are the two major load disturbances that must be
rejected by the control system. Feedback control based on the impurity levels in the product
streams is usually not acceptable due to the large delays / lags introduced by composition
measurements. Also, control action would only be taken after the product purity deviates in a
feedback system. Considering that the column consists of an array of trays, the trays would
respond to a load disturbance much before the effect of the disturbance reaches the product
streams. It therefore makes sense to control an appropriate tray process variable so that the

44
disturbance is compensated for before the product purities are affected. This would lead to
tighter product purity control.

(a) L/D-Q PC (b) L/D-B PC

Condeser Condeser

Reflux Reflux
LC LC
FC FC
Distillate Distillate
FC FC
Feed Feed
L/D L/D
FC
Reboiler steam

LC Reboiler steam LC FC
Bottoms
Bottoms

(c) D-Q/B PC

Condeser

LC

FC
FC Reflux

Distillate
Q/B
Feed

FC

LC Reboiler steam

Bottoms

Figure 5.3. Schematics of L/D-Q, L/D-B, and D-Q/B control structures.

The tray temperature is almost always used as an inferential variable for the tray
composition. The boiling point of a mixture depends on the pressure and the mixture
composition. At a constant pressure, the boiling point increases as the concentration of

45
heavies increases and vice versa. A change in the tray composition can thus be inferred from
a change in the tray temperature. The relationship is exact for a binary mixture and is
approximate for a multi-component mixture.

5.3.1. Single-Ended Temperature Control

Controlling a single tray temperature in a column is usually referred to as single-

ended temperature control. Either of the two operation degrees of freedom can be used as the
manipulation handle. For example, in the LQ control structure, the reflux rate or the reboiler
duty can be manipulated for maintaining a tray temperature. This is shown in Figure 5.4(a).
Manipulation of Q is usually preferred due to the fast response of all the tray temperatures to
a change in Q. The dynamics with respect to reflux rate are slower due to the associated tray
liquid hydraulic lags. The reflux rate can also be used for maintaining a tray temperature if
the control tray location is not too far below the reflux (say about 10 trays). The single ended
variants for the DQ and LB control structures are shown in Figure 5.4(b) and (c) respectively.
In the DQ structure, if the distillate rate is used to control a tray temperature, the temperature
controller is nested with the reflux drum level controller. This means that the temperature
controller would work only if the reflux drum level controller is working. Similarly, in the
LB structure, if the bottoms flow rate controls a tray temperature, the temperature controller
is nested with the reboiler level controller. In both these cases, the level controller must be a
tightly tuned PI controller, else the temperature control would be extremely sluggish. Note
that the reflux and the reboil are the only two causal variables that affect the tray temperature
so that any control scheme must directly / indirectly effect a change in these causal variables.

5.3.2. Dual-Ended Temperature Control

Theoretically speaking, since the column degree of freedom is two, two tray
temperatures can be controlled in a column. This is referred to as dual-ended temperature
control. For example, in the LQ control structure, the reflux rate can be used for controlling a
rectifying tray temperature and the reboiler duty can be used to control a stripping tray
temperature as in Figure 5.5. Industrial practice is to control a single tray temperature as
controlling two tray temperatures usually requires detuning of the temperature controllers due
to interaction between the temperature loops. More importantly, the two controlled tray
temperatures may not be sufficiently independent so that, in the worst case, the control
system may seek infeasible temperature set-points. Dual temperature control is possible for
long towers so that two tray temperatures that are far apart are sufficiently independent.

5.4. Temperature Sensor Location Selection

Various criteria have evolved for the selection of the most appropriate tray location(s)
for temperature control. Prominent among these are selection of tray with the maximum slope
in the temperature profile, sensitivity analysis and SVD analysis.

5.4.1. Maximum Slope Criterion

The maximum slope criterion is the simplest to use and requires only the steady state
temperature profile. From the temperature profile, the tray location where the temperature
changes the most from one tray to the other is chosen as the control tray. The temperature
profile usually also shows a large change / break at the feed tray location. The feed tray
should however not be chosen for control as the changes in temperature would be due to

46
changes in the feed composition / temperature and not due to a change in the separation. A
large change in the temperature from one tray to the other reflects large separation between
the key components so that disturbances in the separation would affect this separation zone
much more than other locations. It therefore makes sense to place the temperature sensor at
that location.

(a)
LQT PC LTQ PC

Condenser Condenser

FC L

Reflux
LC LC
FC Reflux FC
TC

T Distillate Distillate
Feed T

TC Q FC

LC Reboiler steam LC Reboiler steam

Bottoms Bottoms

(b)

DQT PC DTQ PC

Condenser Condenser

LC D LC

FC
FC Reflux FC Reflux
Distillate

T Distillate TC
Feed Feed
T
TC Q FC

LC LC Reboiler steam
Reboiler
Bottoms steam Bottoms

Figure 5.4 Single ended temperature control structures using LQ and DQ scheme

47
 (c)
LTB PC LBT PC

Condenser Condenser

L
FC
Reflux
LC LC
FC FC Reflux
TC
Distillate Distillate
Feed T
T
TC
Reboiler Reboiler
steam steam

LC FC LC
B

Bottoms Bottoms

(d)

DTB PC DBT PC

Condenser Condenser

LC LC D

FC
FC Reflux FC Reflux
Distillate
TC
T Distillate
Feed Feed
T
TC
Reboiler steam Reboiler
steam

LC FC LC
B

Bottoms Bottoms

Figure 5.4 Single ended temperature control structures using LB and DB scheme

48
 (a) LQ PC (b) DQ PC

Condenser Condeser

LC
Reflux
LC
FC FC Reflux
TC Distillate
TR
Distillate TC
Feed TS Feed TR
TC TC
TS

LC Reboiler steam LC Reboiler

Bottoms Bottoms steam

(c) LB PC (d) DB PC

Condeser Condeser

LC
Reflux
LC
FC FC Reflux
TC Distillate
TR Distillate
TC
Feed TS Feed
TR
TC TC TS
Reboiler Reboiler
steam steam

LC LC

Bottoms Bottoms

Figure 5.5. Dual ended temperature control structures using LQ, DQ, LB and DB schemes.

49
5.4.2. Maximum Sensitivity Criterion

Sensitivity analysis recommends controlling the tray with maximum sensitivity to the
control input. The causal variables that effect a change in the tray temperature are the reflux
rate (or ratio) and the reboiler duty. The sensitivity of the ith tray temperature to the reflux
rate (L) and reboiler duty (Q) is defined as
∂T
S iL = i
∂L
and
∂T
S iQ = i
∂Q
respectively. Controlling the most sensitive tray location provides muscle to the controller as
a smaller change in the manipulated variable is needed to bring the deviating temperature
back to its set-point. The open loop steady state gain is large so that a low controller gain
suffices. The low controller gain mitigates sensor noise amplification. Also, a small bias in
the temperature sensor can be tolerated. Plotting the sensitivity of all the trays with respect to
Q and L would reveal the most sensitive tray location. In case two distinct regions of high
sensitivity are observed, dual temperature control should be possible. If not, dual temperature
control is likely to result in the two temperature controllers fighting each other.

5.4.3. SVD Criterion

The SVD analysis is another useful technique for selecting the tray temperature
locations. The sensitivity matrix
S = [SL SQ]
where SL and SQ are column vectors of tray sensitivities, is decomposed using the singular
value decomposition (SVD) as
S = UΣVT.
In the above U and V are orthogonal matrices with the columns constituting the left singular
and right singular vectors, respectively. The Σ matrix is a diagonal matrix. A plot of the first
two left singular vectors (first two columns of U) shows the two most independent locations
in the column. The index of the element of the first left singular with the maximum
magnitude corresponds to the tray location that should be controlled in a single-ended
scheme. If dual temperature control is to be implemented, the corresponding index for the
second left singular vector gives the tray location for the second temperature sensor. The
feasibility of dual ended temperature control is reflected in the ratio of the two diagonal
elements, σ1 and σ2, of Σ. The diagonal elements in Σ are always in decreasing order of
magnitude. If the two singular values are comparable, ie the ratio σ1/ σ2 is not too large (say <
10), dual temperature control should be possible.
Of the above three criteria, the maximum slope criteria is the simplest to use.
Sensitivity analysis and SVD analysis requires the availability of a rating program to
calculate the tray temperature sensitivities to the manipulated variables. The SVD technique
further requires a module to obtain the SVD decomposition of the sensitivity matrix S. In
most distillation column studies, the three techniques would agree on sensor location.
However, for columns with highly non-ideal columns, the use of the SVD technique is
recommended for selecting the tray location.

50
5.5. Considerations in Temperature Inferential Control

5.5.1. Effect of LLK / HHK

The temperature composition relationship is not exact for multi-component mixtures.

If the feed LLK composition increases, the LLK must leave up the top of the column. For the
same LK/HK split, the tray temperatures in the enriching section should be lower as the LLK
composition must increase due to the increase in its feed composition. This dip in the
temperature would be more as one moves up the enriching section since the LLK
accumulates at the top. If a tray temperature near the top is controlled using the reboiler duty
and the tray temperature set-point is not reduced on increase of feed LLK, more of the HK
would be pushed up the top of the column by the action of the controller. Controlling a tray
temperature near the feed tray would mitigate this effect. Another option is to measure the
HK composition in distillate and use it to compensate the control tray temperature set-point.
Note that the volatilities dictate that any LLK entering the column must exit up the top. In
case LLK in the distillate is not acceptable, action must be taken upstream to ensure LLKs do
not enter the column. Troubleshooting the process would typically reveal an upstream
column not doing its job.
Similar to LLK, if a stripper tray temperature low down the column is being
controlled using boilup, an increase in feed HHK would cause more LK to leak down the
bottoms, unless the tray temperature setpoint is appropriately increased.

5.5.2. Flat Temperature Profiles

When the key components in a mixture are close boiling, the column temperature
profile is flat with only a small change in adjacent tray temperatures. This is typical of
superfractionators that use a large number of trays and a high reflux ratio as the separation is
inherently difficult. Controlling a tray temperature is then not desirable as variations in the
tray pressure with changes in column internal flow rates would swamp any subtle variations
in the tray temperature due to composition changes. Controlling the difference in two tray
temperatures that are located close by mitigates the effect of pressure variation as the change
in the local pressure for the two trays would be about the same. The differential temperature
measurement then reflects the change in the HK (or LK) composition between the trays. Care
must be exercised in the use of a differential temperature measurement as the variation in ∆T
with the bottoms composition depends on the location of the separation zone inside the
column. If the measurement trays are below the separation zone, ∆T increases as the steam
rate is decreases. Once the separation zone passes below the ∆T trays, a decrease in the steam
would cause the ∆T to decrease. The gain thus changes sign depending on the location of the
separation zone inside the column.

5.5.3. Easy Separations

The other extreme to a flat temperature profile is an extremely sharp temperature

profile. This occurs when the separation is very easy so that the separation zone shows a large
change in temperature over a few trays ie a sharp temperature profile. During transients, this
sharp separation zone may move up or down the column leading to temperature transmitter
saturation. Once the separation zone moves up and continues to move up, the error signal that
the controller sees does not change so that the burden of bringing the profile back falls on the
integral action. The problem is compounded by the low controller gain due to the extreme

51
sensitivity of the tray temperature to a change in the manipulated variable. The problem is
solved by controlling the average temperature of the trays over which the profile moves.

5.6. Control of Complex Column Configurations

5.6.1. Side-draw Columns

Side product streams are sometimes withdrawn from a column when the product
purity specifications are not very tight and there is small amount of impurity in the feed that
must be purged. Two common configurations are a liquid side-draw from a tray above the
feed tray or a vapour side draw below the feed tray. Consider an ABC ternary mixture. If the
component flow rate of A in the fresh feed is small, the liquid side stream withdrawal above
the feed tray allows most of the B to be removed in the side stream. The side-draw must be
liquid as A being the LLK would be present in smaller amounts in the liquid phase. The
vapour side draw below the feed tray is used when there is a small amount of C (compared to
A and B) in the fresh feed. The C HHK would separate into the liquid phase so that a vapour
side stream that is mostly B with small amounts of C can be withdrawn below the feed tray.
The side stream (liquid or vapour) provides an additional opearation degree of freedom and
its flow rate may be adjusted to maintain the B purity in the side draw. The control schemes
are illustrated in Figure 5.6.
Alternative simpler control schemes are possible when the light A or heavy B
impurities occur in very small amounts in the fresh feed. The purge rate is flow controlled
with a set-point corresponding to the maximum expected impurity component flow in the
feed. When the impurity is below this maximum, small amount of LK or HK would be lost
with purge. However, the loss is acceptable due to the very small purge rate. The alternative
simpler control schemes for the two common side draw configurations are shown in Figure
5.7.

5.6.2. Side Rectifier / Side Stripper Columns

The side rectifier and side stripper columns are an extension of the side-draw column
discussed above. As with side-stream columns, these are used when there is a small amount
of light or heavy impurity in the feed that is removed as a small purge stream. However the
purity specs on the main products are tight. The vapour or liquid side stream respectively,
must then be further rectified or stripped to ensure that the impurity is pushed back into the
main column and does not escape with side-product stream to ensure high purity. The side
stripper and side rectifier column arrangements are shown in Figure 5.8(a) and (b). An
additional operation degree of freedom is introduced in the form of the reflux rate or the
reboiler duty. The side draw rate and the reflux rate or reboiler duty can then be adjusted to
maintain the two impurities in the side-product. The corresponding control schemes are
shown in Figure 5.8(a). Along with the two composition loops in the main column, these
schemes represent a highly coupled 4X4 multivariable system. Simpler control schemes with
only one temperature (or composition) being controlled in each of the main column and the
side-column are much more practical.

52
 Liquid sidestream Vapor sidestream
Condenser Condenser

Reflux Reflux
A A
Distillate Feed Distillate
ABC
B B
Side stream Side stream
Feed
ABC

Reboiler Duty Reboiler Duty

C C
Bottoms Bottoms

PC

Condenser

CC

LC
Reflux
D xD, B
Distillate
Feed
CC
FC S xS, B
Side stream

CC

LC Reboiler Duty

B xB, B
Bottoms

Figure 5.6. Control of side stream column

53
 (a) PC (b) PC

FC
LC
L
LC
FT FC FC L

Ratio Light D D
FC Purge TC
FC F

S S
F

TC

LC Q LC Q

Heavy
FC B Purge
B

F Feed
B Bottoms
D Distillate
Q Reboiler Duty
L Reflux

Figure 5.7. Control of purge columns of (a) Liquid side draw and (b)Vapor side draw

54
 PC

CC

LC
L
FC

D
F
CC

LC CC

CC
Q
LC
B

Figure. 5.8(a). Side stream column with stripper

55
 PC

CC

LC
L
FC
D

CC PC
F CC

LC

LC

CC
Q
LC

Figure 5.8(b). Side stream column with rectifier

56
 PC
PC

FC

LC
LC
FC L
xD2, A
D
F CC
xS2, B
CC

xB1, A
CC

xB2, B
CC
LC
Q

LC B

Figure 5.9. Side stream column with prefractionator

5.7. Control of Heat Integrated Columns

Heat integration arrangements in columns consist of the hot vapour from high
pressure column providing the energy for reboil in a low pressure. The reboiler for the low
pressure column then also acts as the condenser for the high pressure column. Three possible
heat integration schemes are shown in Figure 5.10. In the feed split scheme (Figure 5.10(a)),
a binary fresh feed is split and fed to two columns. One of the columns is operated at high
pressure and the other at low pressure. The pressure difference is chosen so that the hot
vapour is 10-15 C hotter than the low pressure column reboiler temperature. The temperature
difference provides the driving force for reboiling the liquid in the low pressure column. In
the control structure shown, note that the feed to the low pressure column is adjusted so that
the bottoms composition is maintained. This is because the reboiler duty in the low pressure
column cannot be manipulated. Heat integration thus leads to the loss of a control degree of
freedom. Also note that this heat integration scheme is used for a binary separation as the
presence of LLK / HHK components can affect the column temperature profiles sufficiently
so that the temperature driving force necessary for heat transfer in the low pressure column
reboiler disappears.

57
 (a) PC

LC
Feed

LP

CC
CC
CC

HP

LC B2

CC LC

D1

LC
B1

(b)
PC

CC
LC
LP CC
HP
D1

Feed
CC

LC
LC
D2 LC

B1 B2

58
 (c)
PC

CC
LC CC
LP CC
HP
D1

S2
Feed
CC

LC
LC
D2 LC
B1
B2

Figure. 5.10. Heat integrated columns (a) Split of feed (binary); (b) Light split reverse
(binary); (c) Prefractionator reverse (ternary)

Figure 5.10(b) shows the reverse light split heat integration scheme. Approximately, half the
light component is removed as the distillate from the first low pressure column. The bottoms
is fed to the high pressure column to remove the light and heavy components as the distillate
and bottoms respectively. The hot vapour from the top of this column is condensed to provide
vapour reboil into the low pressure column. The direction of heat integration is reverse to that
of the process flow. Hence the name, reverse light split. The forward light split configuration
is also possible. The control structure for this heat integration scheme is self-explanatory.
Figure 5.10(c) shows another heat integration scheme that can be used for ternary
mixtures. The scheme is the same as prefractionator side-draw complex column discussed
previously (Figure 5.9) except for reboiler in the pre-fractionator (low pressure column)
acting as the condenser for the main column (high pressure) through heat integration. The
control structure for this configuration is again self explanatory.

5.8. Control of Homogenous Extractive Distillation System

Homogenous extractive distillation is used to separate a binary AB mixture that cannot be

separated due to relative volatility approaching one or the presence of a binary azeotrope. As
shown in Figure 5.11, A heavy solvent S is added near the top of the first column, the
extractive column, to soak in one of the components (say B). The distillate from the first
column is then near pure A. The bottoms, a mixture of S and B, are fed to the solvent
recovery column that recovers the heavy S in the bottoms and recycles it back to the
extractive distillation column. The distillate from the solvent recovery column is pure B. The
control scheme depicted in the Figure manipulates the reboiler duty in the two columns to

59
keep respectively A and B from falling down the bottoms in the two columns. The solvent
into the extractive column is ratioed to ensure fresh feed rate to ensure enough solvent for
extracting component B. Note that the bottom sump level is not controlled and the sump must
provide enough surge capacity for handling the expected variation in the fresh feed flow rate.
Any loss of the solvent over long time is made up by a make-up solvent stream (not shown).

Solvent PC
PC

FC FC
FC
Ratio LC
A HP LC
LP
FT B

Feed
A&B TC TC

LC

Solvent Extractive Column Solvent Recovery Column

Figure 5.11. Control of Extractive distillation column

5.9. Plant-wide Considerations

In the plant-wide context, the distillate and / or bottoms would feed into a downstream
unit such as another distillation column. The variation in the distillate / bottoms flow rate then
acts as a disturbance into the downstream unit. The LQ control structure is particularly
preferable as the reflux drum and reboiler levels are controlled using the P only controller
which results in a smooth flow change into the downstream unit. If however, the DQ (or LB)
control structure is used in a high reflux ratio (reboil ratio) column and a tray temperature is
controlled using D, the reflux drum level controller manipulates L and must be tightly tuned
for a fast dynamics of the of closed loop temperature controller. The D would then show
large changes disturbing the downstream unit. Feedforward control action can and should be
used to mitigate the propagation of variability to downstream units. For example, the
distillate rate may be ratioed to the feed rate with the composition / temperature controller
setting the ratio set-point. Alternatively, the distillate may be moved in ratio with reflux with
the composition controller setting the reflux ratio set-point. The reflux level controller is then
tuned as a P only controller for smooth changes in the reflux and hence the distillate. The

60
variability in the distillate rate can thus be greatly reduced improving the overall plant-wide
control performance.
The vapour distillate from a partial condenser can be manipulated to control the tower
pressure. This is however not a good idea if the vapour stream feeds directly into a
downstream unit and not a surge tank. In such a scenario, the column pressure should be
controlled using the condenser cooling duty, the reflux drum level controlled using the reflux
rate and the vent rate moved in ratio with the reflux. This arrangement mitigates the
propagation of variability downsteam.
The control of energy integrated distillation columns can also be problematic as a
disturbance on the hot vapour side necessarily affects the boil-up in the reboiler using the hot
vapour as the heat source (instead of steam). To maintain the control tray temperature in the
heat integrated column, an auxiliary reboiler (or condenser, as appropriate) is provided. The
heat integrated reboiler and the auxiliary reboiler may be arranged in a parallel or a series
arrangement. The series arrangement is preferred as the temperature variations in the hot
vapour are attenuated due to variation in the temperature driving force in the auxiliary
reboiler as shown in Figure 5.12(a). In the parallel arrangement, the auxiliary reboiler must
adjust for the variability on the hot vapour side after it has entered the column. One way to
prevent this is to use a total heat input controller as shown in Figure 5.13.

61
 (a)

TC
Orifice TC

PC
set
LC

BFW

FORCED CIRCULATION

(b)

TC

PC
set
TC LC

BFW

Figure 5.12. Reactor/column heat integration with auxiliary reboiler in (a) series (b) parallel

62
 Σ FT
TC

PC
set
set
QC TC LC

BFW
FT

Figure 5.13 Reactor/column heat integration with auxiliary reboiler in parallel and Q controller

63
 Chapter 6. Control of Reactors

A reactor is the heart of a chemical process where the reactants undergo the desired
chemical transformation to the products. The transformation is usually incomplete and is also
accompanied by undesirable transformation to by-products through side reactions. The
reactor operating conditions of temperature, concentration and flow rates determine the
production rate of the main product and the side-products. The downstream separation load
for separating the unreacted reactants from the products and recycling them back to the
reactor is also determined by the reactor. Proper operation and control of the reactor is then
crucial to determining the overall process operating profit. The reactor conversion (or yield)
and selectivity are the two most commonly used reactor performance metrics that are directly
related to the economics of the process. The conversion is defined as the fraction (or %age)
of reactant in the feed that reacts to form product(s). The yield is the conversion of a key
reactant, usually the limiting reactant. The selectivity is defined as the desired product
generation rate relative to the total product generation rate (including all undesired side
products). The yield and selectivity represent two key economic objectives of any process. A
lower conversion would result in greater energy consumption to separate the unreacted
reactants from the products and recycle them back to the reactor. The energy cost per kg
product would then go up. A low selectivity represents an economic loss as more of the
costly reactant gets converted to the undesired product with lower (or worse, a negative)
profit margin. Typically, as the conversion increases, the selectivity decreases so that the
usual philosophy is to operate the reactor near the maximum conversion for which the
selectivity is acceptable (say >95%).
Even as proper reactor operation is the key to profitability, controlling a reactor offers
unique challenges as most reactions are accompanied by the generation or consumption of
heat. The reaction heat generation / consumption alters the temperature of the reaction
mixture which in turn affects the reaction rate and hence the rate of heat generation /
consumption. The coupling of the thermal and reaction effects leads to a highly coupled non-
linear system. The reaction rate, r (kmol.s-1.m-3 or kmol.s-1.kg-1 catalyst), for an irreversible
reaction A + B → C would generally vary as
r = k.cAα.cBβ.
In the above expression, cA and cB are the concentrations (kmol.m-3) of A and B respectively,
k is the reaction rate constant, and α and β are the reaction order (typically > 0) with respect
to A and B respectively. The units of the reaction rate constant depend on the reaction order
and it follows the Arrhenious temperature dependence as
k = k0.e(-E/RT)
where E is the activation energy and k0 is the Arrhenius frequency factor.
The form of the kinetic expressions above shows that the reactant concentration and
the reaction temperature are the two basic manipulation handles for adjusting rate of product
generation in a reactor. The reaction rate doubling for every 10 deg C increase in the
temperature is an oft quoted rule of thumb. The reactor temperature is thus a dominant
variable that significantly affects the reaction rate. When the reactant concentration is
adjusted for changing the production rate, altering a key reactant concentration affects the
reaction rate more than changing the concentration of other reactants. This is because most
reactors must be operated such that one of the reactants is limiting, i.e. in excess of other
reactants. The non-stoichiometric environment is necessary to suppress side reactions. For
example, consider the main irreversible reaction A + B → C. The product C can further react
irreversibly with B to form an undesired product D as C + B → D. For this reaction scheme,
the reactor must be operated in an excess A environment so that the limited availability of B

64
for further participation in the side-reaction suppresses by-product generation. Clearly,
changing the limiting reactant B concentration would affect the reaction rate more than
changing the excess reactant A concentration. In many industrial reactors, the reactor
temperature and the limiting reactant concentration are the two dominant variables that are
directly / indirectly adjusted for changing the product generation rate.
In exothermic reactions, the use of a selective catalyst lowers the activation energy for
the main reaction. The activation energy for the side reactions is thus more than for the main
reaction. In case the temperature is increased for increasing the production rate, the Arrhenius
temperature dependence of the rate constant causes a larger relative change in the side-
reaction rate. Thus if the main reaction rate increases by say 5%, the side reaction rate would
increase by more than 5% (say ~10%). The reaction selectivity thus goes down. The
adjustment of the reactor temperature for increasing the production rate thus must consider
the detrimental effect on selectivity. In many industrial reactors, the reactor temperature is
usually adjusted to compensate for catalyst poisoning / deactivation so that the overall
reaction rates do not decrease over time.

6.1. Basic Reactor Types

The continuous stirred tank reactor, the plug flow reactor and the packed reactor are
the most common reactor types used in the continuous process industry. These basic reactor
types are shown in Figure 6.1. The PFR and PBR are similar except that the latter holds a
catalyst bed to facilitate the reaction. The CSTR and the PFR (or PBR) differ fundamentally
in terms of back mixing. In the PFR (and PBR), the fluid travels along a pipe as a plug so that
every atom entering the reactor spends the same amount of time inside the reactor before
exiting. This time is also referred to as the reactor residence time. Plug flow, by definition,
implies no back mixing. The exact opposite of plug flow is perfect back mixing as in the
CSTR. The back mixing is accomplished using agitators, spargers and fluidization.

CSTR
Reactants in PFR

Reactants in Products out

Products out PBR

Coolant in

Reactants in Products out

Coolant out

Figure 6.1. Basic reactor types

65
6.2. Plug Flow Reactor

6.2.1. PFR Basics

To understand the behaviour of a PFR (or PBR), imagine a plug of fluid flowing
through the reactor. As it moves through, the concentration of the reactants goes down as
they undergo reaction. Assuming adiabatic operation and an exothermic reaction, the heat
released due to reaction would heat up the plug. The increase in the plug temperature causes
the reaction rate to increase further so that the temperature in the initial part of the reactor
increases exponentially. At a sufficient length down the reactor, the reaction rate begins to
decrease due to the limited availability of reactants. For a large enough reactor length, the
reaction rate would go to zero as the limiting reactant gets exhausted. The temperature profile
for an adiabatic PFR thus resembles a sigmoid as shown in Figure 6.2. The difference in the
inlet and outlet temperature is referred to as the adiabatic temperature rise. If the reaction is
highly exothermic, the adiabatic temperature rise is large, which is usually unacceptable due
to reasons such as promotion of side reactions at the higher temperatures, possibility of
catalyst sintering in a PBR, increase in the material of construction cost etc. The cooled PFR /
PBR is then used

Reactor length

Figure 6.2. Temperature profile for an adiabatic PFR

The most common cooled PBR arrangement is shown in Figure 6.3. Catalyst is loaded
in the tubes of a shell and tube heat exchanger. Pressurized water is re-circulated on the shell
side. The water carries the reaction heat to form steam in the steam drum. The high
recirculation rate ensures a near constant temperature on the shell side. Unlike adiabatic
operation, the temperature profile initially increases as the rate of heat generation is more
than the cooling rate and later decreases with the reaction rate decreasing due to reactant
depletion and the cooling rate increasing due to the higher temperature driving force. The
temperature profile thus exhibits a maximum, also referred to as the hot spot. For highly
exothermic systems, the reactor temperature profile can be extremely sensitive to the
operating conditions, in particular the reactor inlet temperature and the shell side temperature.

66
 Steam

BFW
Coolant in

Reactants in Products out

Cooling out

Figure 6.3. Cooled Packed bed reactor

If the reactant inlet temperature (or coolant temperature) is too low, reaction may not
kick in leading to the quenched state. If the reactant inlet temperature is too high, the reaction
can proceed so fast that only a small fraction of the heat released gets removed resulting in a
temperature run-away. The quenched, hot-spot and run-away reactor temperature profiles are
illustrated in Figure 6.4.

Runaway Condition
Temperature

Hot Spot

Quenched state

Axial position

Figure 6.4 Cooled plug flow reactor temperature profiles

67
 The adiabatic or cooled tubular reactor is commonly used in the process industry for
gas phase reactions. The reactor is usually operated at the maximum equipment pressure so
that the reactant partial pressures are as high as possible for maximum reaction rate. This also
reduces the design volume of the reactor for a given conversion. The control of PFRs is
considered next.

6.2.2. Control of PFRs

6.2.2.1. Adiabatic PFR

The adiabatic PFR is the simplest reactor configuration and is used when the adiabatic
temperature rise is acceptable. The reactants (fresh + recycled) are heated to the reaction
temperature using a furnace. The furnace heat duty holds the reactor inlet temperature
constant. This is shown in Figure 6.5(a). The adiabatic temperature rise sets the reactor outlet
temperature. Sometimes the outlet temperature is also controlled to maintain the reactor
conversion and selectivity. The limiting reactant fresh feed rate may be used as the
manipulation handle as illustrated in Figure 6.5(b).

(a)
Recycle A

Reactor
A Products

B TC

Limiting
Reactant Furnace
Heat duty

(b)

Recycle A Furnace

Reactor

Pure A Products

TC
Pure B
Limiting FC TC
Reactant Set

Figure 6.5. Simple adiabatic PFR structures

68
6.2.2.2. Cooled Tubular Reactors

Cooled tubular reactors are more challenging from the control perspective. The hot spot
temperature must be tightly controlled to prevent a runaway. This is accomplished using
auctioneering temperature control as illustrated in Figure 6.6. The measurements from an
array of thermocouples (or RTDs) placed along the length of the reactor are input to a high
selector that passes the maximum temperature to the hot spot temperature controller. This
controller typically manipulates the reactor cooling duty. The control structure for the most
common cooled PBR arrangement is shown in Figure 6.6. The temperature controller sets the
steam drum pressure set-point. A change in the drum pressure alters pressurized water boiling
point which in turn changes the temperature driving force for heat removal from reactor.

Set
PC Steam
LC

Boiler feed
water
Recycle A Furnace
Reactor

Pure A
Products
TC
Pure B
Limiting FC
TC HS
Reactant

Figure 6.6. Auctioneering temperature control structure

In addition to the reactor cooling duty, there are two other possible manipulation handles
for reactor heat management, namely the reactor inlet temperature set-point and the limiting
reactant flow rate into the reactor. The schemes are shown in Figure 6.7. Both the schemes
work by changing the heat generation due to reaction. The non-linearity between the
controlled and manipulated variable is severe in all the three schemes. This may be
understood using the analogy of a fire. It is very easy to make a fire whereas extinguishing
one requires much effort. Similarly, it requires much more control effort to adjust for an
increase in the hot-spot temperature than for the same decrease. The controller thus must be
aggressive for deviations in one direction and not too aggressive in the other. Gain scheduling
is sometimes used with the magnitude of the controller gain depending on the magnitude and
sign of the error signal (or a more sophisticated schedule). The possibility of temperature run-
away also requires that large overshoots above the set-point be avoided even as the controller
is aggressive. The derivative action is often employed to suppress closed loop oscillations.

69
 (a)
PC Steam
LC

Boiler feed
water
Furnace
Recycle A Reactor

Products

Pure A
TC
Pure B
Limiting Set
FC HS
Reactant TC

(b)

PC Steam
LC
Boiler feed
water

Furnace
Recycle A Reactor

Pure A Products
TC
Pure B
Limiting
Reactant FC TC HS
Set

Figure 6.7. Two other possible manipulation handles for reactor heat management

It is noted that controlling the reactor outlet temperature may sometimes provide
adequate regulation of the hot spot temperature. The applicability of this much simpler
scheme depends on how close the outlet is to the reactor hot spot at the base design condition.
When the hot spot is close to the reactor outlet, a change in the outlet temperature correlates
well with the change in the hot spot temperature so that controlling the outlet temperature
provides adequate regulation of the hot spot temperature.

70
6.2.3. Intermediate Cooling and Cold–Shot Cooled Reactors

Two other commonly used heat removal configurations, namely, intermediate cooling and
cold shot cooling, are shown in Figure 6.8 along with the control structure. Explicit
intermediate coolers are provided in equilibrium limited exothermic reactions to increase the
overall equilibrium conversion. Cold shot cooling is frequently employed in polymerization
reactors where it is extremely important to hold the temperature profile in the reactor to
maintain the molecular weight and polydispersity of the product polymer.

TT TT

TC TC

(a)

TT TT

TC TC

(b)
Figure 6.8. (a) Intermediate cooling in sequence reactors
(b) Cold shot cooling

Modern chemical plants frequently recover the reaction heat using heat integration. A
common heat integration scheme employs the feed effluent heat exchanger to preheat the
cold reactor feed using the hot reactor effluent gases as in Figure 6.9. The heat integration
results in energy recycle loop which can lead to instability. Consider the extreme case of a
FEHE that heats the reactants to the reaction temperature so that there is no furnace. If the
temperature of the cold reactants rises, the reactor inlet temperature will increase. This would
cause more reaction accompanied by heat release resulting in an increase in the hot reactor
effluent gas. The hotter effluent would cause a further increase in the reactor inlet
temperature resulting in a temperature runaway or instability. This instability can be
prevented if the reactor inlet temperature (or outlet temperature) is controlled. The furnace
performs this function by breaking the positive thermal feedback loop. An alternative to the
FEHE is to recover the reaction heat as steam in a waste heat boiler and feed it to the steam
utility network. This removes the thermal feedback while ensuring almost 100% heat
recovery.

71
 Energy recycle

Furnace
Reactor

FEHE
Cold
reactor Products
feed
TC
Products to
separation section
Fuel

Figure 6.9. Heat integration scheme

Catalytic packed bed reactors differ from simple plug flow reactors in that that the
catalyst bed constitutes a significant thermal capacitance. The heat capacity of the packed bed
can sometimes lead to the outlet temperature exhibiting an inverse response with respect to
the inlet temperature. This effect is due to the difference in the propagation rate of the gas and
the bed thermal effect through the reactor. If the reactor inlet temperature decreases, the bed
temperature does not decrease immediately. The cooler gas plug thus comes in contact with
the hot packing and heats up. The reaction now kicks off leading to the outlet temperature
increasing. Once the catalyst bed cools down, the outlet temperature of course decreases. This
inverse response or “wrong” way behaviour destabilizes the control loop so that a PID
controller must be detuned. In cases where the closed loop performance is not satisfactory,
the application of advanced control techniques such as the Smith predictor is recommended.

6.3. Continuous Stirred Tank Reactor

Perfect back-mixing occurs in ideal continuous stirred tank reactors. Due to the
mixing, the composition at the reactor is the same as the composition inside the reactor. The
reaction thus occurs at the reactor outlet concentration. The reactant, upon entering the
reactor, thus gets diluted instantaneously to the lower reactor composition. Since the reaction
occurs at the exit conditions, the steady state conversion now depends on both the inlet and
outlet conditions. This creates higher non-linearity so that the existence of multiple steady
states is a distinct possibility. This is in direct contrast to PFRs where the outlet conditions
are uniquely determined by the inlet conditions. Back-mixing thus causes material feedback
with the possibility of multiple solutions.

72
 Consider a jacketed CSTR as in Figure 6.10. Assume that the coolant flow rate is high
so that the jacket temperature is nearly constant. The heat removal rate varies linearly with
the jacket-reactor temperature difference. The heat generation due to reaction is an S-shaped
sigmoid with respect to the reactor temperature. At steady state the heat removal rate and the
heat generation rate must balance each other. The reactor can exhibit a unique steady state or
multiple steady states. These scenarios are depicted in Figure 6.11. For multiple steady states,
there are three steady states corresponding to a high, intermediate and low temperature. The
high and low temperature steady states are stable while the intermediate temperature steady
state is unstable. This is because around the intermediate steady state, if the temperature
increases slightly, the rate of heat generation increases more rapidly than the rate of heat
removal so that the temperature would continue to rise and not return back to the intermediate
steady state ie an open loop unstable system. In contrast, at the high / low temperature steady
state, the slope of the heat removal curve is more than the heat generation curve in the
vicinity of the steady state implying stable open loop behaviour. Typically, reactor operation
at the intermediate steady state is desired as the high temperature steady state may lead to
catalyst sintering and undesirable side reactions while the reaction rate is small at the low
temperature steady state. For such open loop unstable reactors, a temperature controlled must
be used to stabilize the reactor. The closed loop system becomes stable for a controller gain
above a critical value. At lower gains, the feedback action is not enough to stabilize the
unstable system. For extremely large controller gains, the closed loop system becomes
unstable due to too much feedback, similar to open loop stable processes. The closed loop
system is thus stable only for a range of controller gain. Conventional tuning rules such as the
ZN / TLC procedure should therefore not be applied. Multiple steady states are avoided when
a large heat transfer area is provided. Unique solution CSTRs are much easier to control and
the heat transfer system for CSTRs should be properly designed to avoid multiplicity.

Cooling
Water

Figure 6.10. High flow rate cooling water temperature control

73
 Unique Steady State

Multiple Steady States

Reactor temperature

Figure 6.11. Steady states in CSTR

6.3.1. Jacket Cooled CSTR

Evidently, proper regulation of the reaction heat removal is one of the main control
tasks. The simplest heat transfer arrangement is that of a jacketed CSTR with cooling water
flowing along the jacket. Its flow rate is adjusted to maintain the reactor temperature as
shown in Figure 6.12. This approach suffers from drawbacks such as a non-linear process
gain due to variation in the heat transfer coefficient with cooling water flow and local reactor
hot spots due to the jacket temperature profile. These problems can be mitigated by using a
recirculation loop as shown in Figure 6.13(a). The recirculation allows for a constant coolant
flow rate so that the jacket temperature is constant. The reactor temperature is maintained by
adjusting the fresh coolant flow rate into the recirculation. The recirculation loop introduces
and additional thermal lag into the heat transfer loop resulting in a slow closed loop response.
The use of a cascade control scheme as in Figure 6.13(b), where the jacket temperature is
tightly controlled by adjusting the coolant flow (slave loop) and the reactor temperature loop
adjusts this set-point, significantly improves the closed loop reactor temperature control.
TT TC

Cooling
Water

Figure 6.12. Simplest CSTR Control Structure

74
 TT TC

Cooling
Water

(a)

TT TC

SP
TT TC

Cooling
Water

(b)

Figure 6.13 (a) Circulating cooling water temperature control

(b) Cascade control

6.3.2. Reaction Heat Removal as Steam

A common arrangement for recovering the reaction heat is to generate steam from the
hot pressurized water recirculating in the jacket recirculation loop. The heat removal scheme
and the control structure are shown in Figure 6.14. The level in the steam drum would exhibit
the inverse response so that the boiler feed water flow into the drum is ratioed to the steam
flow with the level controlled adjusting its set-point. This arrangement allows for the feed
water to move in the correct direction for load changes.

75
 SP
TT TC
PT PC LP or HP
Steam

SP
LC Х FI
LT SP
FC

Boiler Feed Water

Figure 6.14. Temperature control through steam generation

6.3.3. External Heat Exchanger

In high temperature applications, the generation of steam is not possible so that

expensive proprietary coolant oils must be used. An external heat exchanger is then provided
as in Figure 6.15 for a closed circuit coolant loop. Cooling water is used as the coolant in the
external heat exchanger to remove the reaction heat carried by the hot oil. The control
structure shown in the Figure is self explanatory.

TT TC

SP

Split ranged valves

TT TC

Cooling water in

Figure 6.15. Use of external Heat exchanger for temperature control

76
6.3.4. Cooling Coils

In the jacketed CSTR, the heat transfer area is determined by the reactor volume and
may not be sufficient. Cooling coils, as in Figure 6.16, are used for higher heat transfer area
per unit volume. The control scheme adjusts the coolant flow rate for maintaining the jacket
temperature.

TT TC

Cooling
Water

Figure 6.16. Extended heat transfer area through use of cooling coils

6.3.5. External Cooling by Content Recirculation

Another alternative for removing the reaction heat when the jacket heat transfer area
is insufficient is to circulate the reaction mixture through an external heat exchanger and feed
it back into the reactor. This scheme is shown in Figure 6.17. The reactor temperature is
maintained by manipulating the temperature of the cooled reaction mixture from the external
heat exchanger. The cooling water flow rate into the heat exchanger is adjusted to maintain
the cooled reaction mixture temperature. In this control scheme, the external heat exchanger
introduces a significant thermal lag into the slave loop. The dynamics of the slave loop can be
significantly improved by a slight design modification providing for bypassing a small
fraction of the reaction mixture stream around the external heat exchanger. This is illustrated
in Figure 6.18. The thermal lag is thus replaced by the negligible mixing lag as the dominant
time constant of the slave loop resulting in significantly improved reactor temperature
control.

6.3.6. Boiling CSTR with External Condenser

When the reaction mixture boils, excellent reactor temperature control can be
achieved using an external condenser that condenses the vapour and refluxes the cold
condensate back into the reactor. The arrangement is shown in Figure 6.19. Note that the
condensate flows back into the reactor by gravity so that the condenser should be at a
sufficient elevation above the reactor. The U-leg seal is provided to force the vapour to enter
the condenser from the correct entry port. Note that the reactor temperature in this case is
self-regulatory..

77
 TT TC

Cooling
Water

Figure 6.17. Circulation of reactor contents through external heat exchanger

TT TC

Set
Split
TC ranged
valves

TT

Cooling
Water

Figure 6.18. Bypass of circulating reactor content around external Heat exchanger

78
 PT
Vapor to
Condenser SP
PC
Liquid
return
to reactor Cooling water in

Liquid height

U-leg seal

TT TC

Figure 6.19. Cooling through vaporization of reactor content

6.3.7. Reactor Heat Removal Capacity Constraint

A possibility for controlling the reactor temperature when the heat transfer area is
limiting is to adjust the reactant feed rate so that the heat generation due to reaction changes
appropriately. The cooling water valve is fully open. The scheme is shown in Figure 6.20.
While, appealing in its own right, this control strategy should not be implemented in practice
(or used with due caution) as the open loop dynamics of the temperature loop is slow due to
the composition lag introduced by the reaction mixture volume. As the feed rate changes, the
composition of the reaction mixture changes slowly due to the large reactor hold up. The
reaction heat generation thus changes slowly. In case the reactor temperature goes down, the
temperature controller would add more feed. The unreacted reactant amount in the reactor
thus goes up. Once the reactor temperature begins to increase, reaction would “kick-in” due
to the large amount of unreacted reactant inside the reactor. The possibility of a reactor run-
away is thus always lurking in the back-drop, especially for highly exothermic reactions. The
scheme may be workable for mildly exothermic reactions.

TC TT

Fresh feed reactant

Feed or catalyst

Cooling Water valve

Wide open

LT LC

Figure 6.20. Temperature control by manipulating the fresh feed rate

79
 The problem of reactor runaway can be circumvented by the use of a valve
positioning scheme as illustrated in Figure 6.21. The reactor temperature is controlled by
adjusting the reactor cooling duty. The valve positioning controller measures the cooling duty
valve position and slowly adjusts the feed rate so that eventually the cooling duty valve is
near fully open and the reactor operates at maximum through-put. In this scheme, the
temperature control loop effectively rejects short term disturbances as its dynamics are much
faster compared to controlling the reactor temperature directly using the fresh feed rate. Over
the long term, the VPC ensures the reactor is operating at near maximum cooling duty, ie
maximum through-put. Note that the feed rate can be directly manipualted in a PFR since the
material flows through as a plug and there is no back-mixing implying little / no build-up of
unreacted reactants inside the reactor.

Set
Set VPC
90 %

FC

TT TC

Set
Fresh feed
reactant
TT TC

Cooling
Set Water out

FC

Cooling
Water in

Figure 6.21. Valve positioning control for throughput maximization

80
 Chapter 7. Heat Exchanger Control
Heat exchangers are widely used for heating / cooling process streams to the desired
temperature or to change the phase of a stream. The heat exchanger is thus for removing /
adding sensible or latent heat. Figure 7.1 shows the schematic of a counter-current shell and
tube heat exchanger. The hot stream flows through the tubes and loses its heat to the cold
stram flowing through the shell. The heat exchange is driven by the temperature difference
between the shell side and the tube side. For a given inlet temperature of the hot and cold
streams, the temperature driving force is more for the counter-current flow arrangement.
Most exchangers are thus operated with counter-current flow.

Coolant
in

Hot Hot
stream in stream
out
Coolant
out

Figure 7.1. Counter current Shell and Tube heat exchanger

The heat exchangers in a process can be usefully classified into utility heat
exchangers or process-to-process heat exchangers. Utility heat exchangers typically use
steam or cooling water to respectively add or remove heat from a process stream. In process-
to-process heat exchangers, both the hold and cold streams are process streams.

7.1. Control of Utility Heat Exchangers

The purpose of a utility heat exchanger is to provide (remove) as much heat as is

necessary to maintain the process stream temperature. A simple temperature controller that
adjusts the utility flow rate to maintain the process stream temperature accomplishes this
function. Figure 7.2(a) shows a cooler that uses cooling water for heat removal. Figure 7.2(b)
shows a heater using steam as the utility fluid. The control loops are self-explanatory.
Sometimes the heat transfer is controlled without adjusting the utility flow. For
example in a partial condenser that vents the non-condensables as a vent stream, the cooling
water valve is fully open and the vent rate is adjusted to control the condenser pressure. The
pressure sets the dew point temperature of the condensables in the vapour stream fixing the
temperature driving force across the tubes. The control scheme is illustrated in Figure 7.3.

81
 Coolant
TC
in

TT

Hot Hot
stream stream
in out
Coolant
out
(a)
Steam
TC

TT

Cold Cold
stream stream
in out
Steam
trap
(b)
Figure 7.2. Control of utility exchangers

PT PC

Coolant
water
Inerts

Column Process
vapors stream out

Coolant
out

Figure 7.3. Indirect control of Heat exchanger using partial pressure of non-condensable

The flooded condenser is another common arrangement where the level of the condensate
determines the number of tubes that are submerged. Heat is thus transferred only across the
tubes exposed to the vapour. The cooling rate thus gets adjusted to maintain the pressure by
manipulating the condensate draw which affects the level. The liquid hold up inside the
condenser represents a significant lag (~2-5 minutes) so that the pressure cannot be controlled
very tightly. Flooded condenser arrangement is shown in figure 7.4.

82
 Column Vapors

PT

Coolant
in

PC

Figure 7.4. Heat transfer control by variable heat transfer area in flooded condenser

7.2. Control of Process-to-Process Heat Exchangers

Process to process heat exchangers transfer heat between two process streams. The
flow of these process streams is usually set elsewhere in the plant so that adjusting the flow
rate of one of the process streams to regulate the amount of heat transferred is not possible.
To provide a control degree-of-freedom for regulating the heat transferred, a small by-pass
(~5-10%) of one of the process streams around the heat exchanger is provided. The outlet
temperature of this process stream or the other process stream can be controlled by
manipulating the by-pass rate. These two schemes are illustrated in Figure 7.5. In the former,
tight temperature control is possible as the amount of heat transferred is governed by the
bypass. In the latter, a thermal lag of the order of 0.5 to 2 minutes exists between the
manipulated and controlled variable.
TC
By pass

Cold
stream

TT

Hot Hot
stream in stream out

Cold stream
out
(a)

83
 TC
By pass

Cold
stream

TT

Hot stream Hot

stream out

Cold stream
out

(b)

Figure 7.5. By-pass control of process to process heat exchangers

(a) Controlling and bypassing hot stream
(b)Controlling cold stream and bypassing hot stream

Process-to-process heat exchangers are increasingly used for heat integration to

minimize the energy consumed per kg product. Given that the flow of the two process
streams into the heat exchangers is set elsewhere, an auxiliary utility heat exchanger is often
provided to control the temperature of the more important process stream. The size of the
auxiliary utility heat exchanger should be large enough for effective disturbance rejection.
In the plant-wide context, heat integration using process-to-process heat exchangers
causes interaction between the interconnected units. In particular, it is necessary to ensure
that in the quest for maximum energy recovery, energy recycle circuits do not lead to
instability. Sufficient control degrees-of-freedom should be provided in the form of auxiliary
utility exchangers so that the variability is transferred to the plant utility system.

84
 Chapter 8. Control of Miscellaneous Systems

In this chapter, the control of other common units in the industry such as furnaces,
compressors, refrigeration systems and plant utility systems is briefly described.

8.1. Furnace Controls

A furnace heats a process stream to high temperature using combustion of a fuel as

the heat source. It consists of a fire box or combustion zone with tube bundles carrying the
process stream to be heated. Fuel is burnt with air in the combustion zone to heat the tubes to
very high temperatures. Typically a convective heat transfer zone is also provided in furnaces
to recover heat from the hot flue gases. The furnace is essentially a reactor combusting fuel
with air. The control objective is to satisfy the ‘on-demand’ heat load. The control system
shown in Figure 8.1 is typically used. The fuel-to-air ratio must be nearly stoichiometric for
complete combustion of the fuel. Excess air is not fed in as that would increase the flue gas
discharge rate. Less than stoichiometric air would lead to partial combustion or worse,
unburnt fuel remaining in furnace. The flue gas oxygen concentration is a good indicator of
the quality of combustion and adjusts the fuel-to-air ratio set-point. The air is fed in ratio to
the fuel. The forced draft fan speed is varied to change the air feed rate. An induced draft fan
is provided at the outlet to suck the flue gases out of the furnace. Its speed is controlled to
maintain the pressure inside the combustion chamber.

O2 C O2 T
Cold Load
Air

FT FC Hot Load
SP

Ratio HS TT

Lag
TC
LS

FT FC
SP
Fuel

Figure 8.1. Furnace firing controls

A critical safety requirement is to operate the furnace such that the air is fed in excess
during transients (load changes). This is necessary to ensure that all the fuel fed into the
furnace is burnt and no unreacted fuel remains inside, lest it combust later to damage the
furnace. Thus if the heat load increases, the air rate must be increased before the fuel valve is
opened. On the other hand, if the heat load decreases, the fuel valve must be closed before the

85
air flow is reduced. This control action is accomplished by lagging the heat load signal as
shown in Figure 8.1. The lagged and the unlagged signals are then input to a high selector
and a low selector. The output of the high selector sets the air flow controller set-point. The
output of the low selector sets the fuel flow controller set-point. If the heat load increases, the
high selector sends the unlagged signal to the air flow controller causing an instantaneous
increase in the air flow. The low selector sends the lagged signal to the fuel flow controller.
The fuel flow thus lags behind the air flow for an increase in the heat demand. For a decrease
in the heat load, the high selector sends the lagged heat load to the air flow controller while
the low selector sends the unlagged heat load to the fuel flow controller. The air thus lags
behind the fuel for a heat load decrease. Furnace operation in excess air is thus ensured
during transients.

8.2. Compressor Controls

Compressors are used to increase the pressure of gas stream. A cooler with a knock-
out pot is typically provided at the compressor outlet to cool the hot pressurized gas and
remove any condensables that liquefy due to the higher pressure and cooling. There are three
important types of compressors used in plants, namely, centrifugal, axial and reciprocating. In
reciprocating compressors, the through-put is adjusted by manipulating the strokes per minute
or the length of a stroke. A recycle is always provided around the outlet of the compressor for
the safety of the compressor.
Centrifugal compressors are similar to centrifugal pumps in that a rotating motor is
used to impart energy to the fluid. To control the through-put, three configurations are
typically used, namely, exit recycle, suction throttling and motor speed manipulation. These
three schemes are illustrated in Figure 8.2. In the exit recycle scheme, a recycle around the
compressor back to the inlet is provided which is adjusted to manipulate the through-put.
Note that the total (recycle + fresh) flow rate through the compressor remains the same so
that compressor operates at a single point on its characteristic curve. This is the most energy
inefficient method of compressor operation. Also, note from the figure that the recycle is
provided after the cooler so that energy recycle is prevented. In suction throttling, a valve is
provided at the compressor suction and the through-put is manipulated by adjusting the
suction pressure. At lower through-puts lesser energy is consumed as the amount of material
flowing through the compressor is less. The most energy efficient method of throughput
manipulation is to vary the rpm of a variable speed drive. High pressure steam, as in a
turbine, is used many a times to provide the motive force for rotation. A cascade speed
controller that adjusts the steam flow rate set-point maintains the drive speed. The drive
speed set-point is input remotely by the through-put controller.
Another important consideration in compressor control is the prevention of surge at low
flow rates. The compressor characteristic curve shows a maximum and the compression ratio
dips a low flow rates due to compressibility. So much so, that if the flow rate goes low
enough, the flow through the compressor can reverse direction. This causes the suction
pressure to build and the flow almost immediately reverses direction again (i.e. flows out the
compressor). This flow reversal cycle repeats in less than a second. To prevent the
compressor surge phenomenon, the compressor discharge is recycled to the compressor
suction. An anti-surge controller, as in Figure 8.3 adjusts the recycle rate to prevent the flow
through the compressor from dropping below a minimum. Note that this minimum must be
sufficiently above the surge flow rate for the particular compressor rpm (or maximum rpm for
variable speed drives).

86
 F Bypass
FC

Suction Motor
P suction
FT

F F Comp
F
Discharge
Compressor P dischg
Coolant
(a)

FC
Motor

FT
Suction

Discharge
Compressor

(b)

FC SC
SP

ST

LP Steam Turbine HP Steam

FT
Suction

Discharge
Compressor

(c)
Figure 8.2. Compressor controls
(a) Exit recycle
(b) Suction throttling
(c) Motor speed manipulation

87
 F Bypass

Motor FminC

FT
Suction
F

Discharge
Compressor
Coolant
Figure 8.3. Compressor antisurge controller

8.3. Decanter Control

Decanters separate a heterogenous liquid-liquid mixture into its constituent liquid

phases by utilizing the density difference between the liquid phases. The heterogenous
mixture phase separates into a heavy and light liquid phase. Typically, the heavy liquid phase
is aqueous while the light liquid phase is organic. Appropriately located withdrawal ports are
provided in a decanter for removing aqueous and organic streams. To prevent the aqueous
liquid from entering the organic liquid withdrawal port, the level of the liquid-liquid interface
must be controlled. Also the organic phase level must be controlled. The simplest scheme,
shown in Figure 8.4(a) manipulates the organic and aqueous stream flow rates to adjust the
respective levels. The organic level controller must however interacts with the aqueous level
controller. A simple and effective strategy for removing the interaction is to adjust the total
flow out from the decanter to control the organic phase level. Figure 8.4(b) shows the
corresponding control scheme. The organic stream flow is manipulated to maintain total flow
(organic + aqueous) out of the decanter. The organic level controller sets the total flow set-
point.

88
 LC

LC
Light liquid phase

LT LC

Heavy liquid phase

(a)

LC

LC
Light liquid phase

LT FT

SP

FT + FC

LC

Heavy liquid phase

(b)

Figure 8.4. Decanter Controls

(a) Conventional level control
(b) Buckley control structure to eliminate interaction

89
8.4. Control of Refrigeration Systems

We study control schemes for the commonly used vapor compression and vapour
absorption refrigeration cycle.

8.4.1. Vapor Compression Cycle

The refrigeration cycle typically employs compression. The cold refrigerant absorbs
heat from the process stream and vaporizes in the evaporator. The vapour is compressed so
that at the higher pressure, cooling water can be used to condense the refrigerant. The
condensed refrigerant is collected in a surge drum and fed to evaporator. Figure 8.5 shows
control schemes for the compression refrigeration cycle. The chilled process stream
temperature controller sets the evaporator pressure set-point. The evaporator pressure is
controlled by adjusting the compressor suction valve. The level in the evaporator is controlled
by adjusting the liquid refrigerant inlet valve. In case a variable speed drive compressor is
used, the pressure controller is done away with and the temperature controller directly sets
the drive speed set-point. The pressure controller is necessitated in the compressor suction
throttling scheme to compensate for the throttling valve non-linearity. In the variable drive
speed compressor, the variation in the suction pressure (evaporator pressure) with respect to
the drive speed is relatively linear so the drive speed can be directly adjusted by the
temperature controller. The level in the refrigerant surge drum is not controlled as the
refrigerant forms a closed circuit. Notice that the heat transfer rate changes as the temperature
driving force across the condenser changes due to the dependence of refrigerant boiling
temperature on the evaporator pressure.

8.4.2. Vapor Absorption Cycle

In addition to compression systems, refrigerant absorption systems are also applied

industrially. The absorption based refrigeration cycle and its control scheme is shown in
Figure 8.6. Ammonia (refrigerant) rich strong liquor is distilled at high pressure to recover
liquid ammonia as the distillate and ammonia lean weak liquor as the bottoms. The liquid
ammonia is fed to the evaporator where it absorbs heat from the process stream to be chilled
and evaporates. Vapor ammonia is absorbed by the ‘weak liquor’ water stream. The ‘strong
liquor’ so formed is fed to the distillation column to completed the closed circuit refrigerant
loop. The temperature of the chilled process stream is controlled by adjusting the level set-
point of the evaporator. The heat transfer rate is thus varied by changing the area across
which heat transfer occurs. The evaporator level controller adjusts the distillate liquid
ammonia flow. An increase in the level of the evaporator implies an increase in the ammonia
evaporation rate so that the weak liquor rate is increased in ratio to absorb the ammonia
vapours. The strong liquor is cooled and collected in a surge drum. The level of the surge
drum is not controlled. Liquid from the surge drum is pumped back to the distillation column
through a process-to-process heater that recovers heat from the hot ‘weak liquor’ bottoms
from the distillation column. The flow rate of the strong liquor to the column is adjusted to
maintain the column bottoms level. Also, the steam to the reboiler is manipulated to maintain
a tray temperature.

90
 SP
TC PC
Compressor

TT

PT Coolant in
Chilled Condenser
stream
out
Evaporator
LT

Hot stream in
LC

Figure 8.5. Compression refrigeration controls

Ammonia vapor
Chilled
PC Stream TC
Out TT
SP
LC
Evaporator
LC

Hot stream in
TC Ammonia Liquid
FT

Ratio FC
SP
LC

Steam LT

Weak Liquor

Strong Liquor

Figure 8.6. Absorption refrigeration controls

91
8.5. Control of Steam Utility System

Figure 8.7 schematically shows a plant power and utility system. Boiler feed water is
heated in a furnace to produce saturated steam. The saturated steam is passed through the
furnace to produce superheated steam at 1000 psia pressure. The superheated LP steam drives
a turbine to produce electricity. Steam at different pressures is extracted form the turbine for
process steam utility requirements. Typically, steam at 300 psia (high pressure steam), 150
psia (medium pressure steam) and 50 psia (low pressure steam) is made available as a heat
source at different temperature levels for process use. The pressure of the superheated steam
from the furnace is maintained by adjusting the furnace duty. The steam drum level is
controlled by adjusting the boiler feed water rate. The pressure of the 300 psia header is
maintained in a split range arrangement as shown in the Figure. Steam from the higher
pressure header is let in for a decrease in the header pressure while steam is dumped to the
lower pressure header for an increase in the header pressure.

Generator
Superheated steam Turbine 300# Generators

300# Header

PC

300# Users
Sat. steam
150# Generators

LC 150# Header
PC
PC
150# Users
BFW
50# Generators
Furnace 50# Header
PC
Fuel
Condensate
Return 50# Users
Vent

Figure 8.7. Power plant utility system controls

92
 MODULE III

ISSUES IN PLANTWIDE CONTROL SYSTEM DESIGN

The control structures for common unit operations as presented in the previous section may give
the impression that developing effective control systems for a complete plant should be a piece
of cake in that we simply put in the control structures for each of the individual unit-operations.
As we will see in this module, there are unique challenges presented by material / energy recycle
that make the plantwide control structure design problem much more challenging than simply
putting in structures for each of the individual unit operations. In fact, there are many-many
reasonable structures that will work to provide safe and stable operation on a given process. The
economic performance of these different structures can however be significantly different.
Industrial examples with prudent altering of the control structure resulting in the maximum
achievable throughput for the same plant increasing by as high as 20-30% are part of industry
folklore. What are the specific plantwide issues that must be considered and addressed in the
design of such effective (including economics) plantwide control systems is the focus of this
module.

For a firm grip on the plantwide control problem, we start from scratch covering degrees of
freedom (control and steady state) and the tremendous flexibility that exists in the choice of the
controlled variables (CVs) corresponding to these dofs as well the combinatorial complexity in
the manipulated variables (MVs) used to regulate these CVs. We also discuss the snowball effect
due to non-linearity caused by material recycle and the integrating nature of the component
inventories in a recycle loop. We then discuss the design of the plantwide regulatory control
system using the conventional CV-MV pairing approach and the more recent, Luyben pairing
approach, along with an illustration on two toy-problems. Finally we bring in economic
considerations and show how these considerations may require operating the plant at or close to
equipment capacity constraints. We also discuss different ways of handling these constraints and
their pros and cons in the plantwide context including illustrations on the two toy examples.

94
 Chapter 9: Control and Steady State Degrees of Freedom

9.1. Control Degrees of Freedom

The plantwide control system design problem can be considered as devising the “best”
strategy for managing the available degrees of freedom (dof) in a process. From the operations
perspective, a degree of freedom may quite simply be interpreted as having the freedom to make
an adjustment, usually to a process / utility flow (a control valve opening). With no control
system on a process, the operator is free to adjust the opening of the available independent
control valves. These are referred to as the control degrees of freedom. By independent control
valves, we imply respecting hydraulic fluid flow laws so that eg on a fluid flow pipe, only a

Figure 9.1. Examples of properly and improperly installed control valves

(a) Flow through a pipe
(b) Flow splitter
(c) Process to process heat exchanger

single control valve is adjusted. Figure 9.1 provides illustrative examples of proper and improper
installation of independent control valves.

How should adjustments be made to the independent control degrees of freedom (control
valves). First and foremost, these adjustments must ensure safe and stable process operation.
This requires using a control system for stabilization of potential instabilities and avoiding
undesirable drifts in process variables. Reactor thermal runaway is an example potential
instability. Process inventories such as liquid levels or gas pressure are examples of process
variables that drift in the absence of proper regulation leading to potentially unsafe situations
such as a tank running dry / overflowing or a rupture disc breaking open to release pressure. The
control system for safe and stable process operation is referred to as the basic regulatory
plantwide control system.
Given basic regulatory control that ensures safety, stability and acceptably small drifts,
further adjustments may be made to any remaining valves or to the setpoints in the regulatory
control system for ensuring the process is operated in the most profitable manner. This may
95
correspond to operating condition adjustments (valve positions or regulatory loop setpoints) to
e.g. minimize steam consumption per kg product, maximize yield to the desired product, on-aim
product quality with no product give-away, proper effluent discharge management etc.

9.2. Steady State Degrees of Freedom

For continuous chemical processes, it is the steady state at (around) which the process is
being operated that determines the operating profit. Of all the control degrees of freedom, not all
affect the steady state. This is illustrated for a very simple 'three-tanks-in-series' process in
Figure 9.2. There are four control valves. Since liquid level in a tank is non-self regulatory (i.e.
unless the inflow and outflow are exactly balanced, the level is either rising or receding), all
three tank levels must be controlled to avoid large drifts in the levels. This would take away three
control valves leaving one valve free. Let us say this free valve is at the process feed. We may
then flow control the feed stream using this valve to set the fresh feed flow at the desired value.
The level controllers then adjust the respective tank outlet valves as shown in Figure 9.2. The
operator can adjust 4 setpoints (one fresh feed flow setpoint and three level setpoints). Of these
the final steady state is determined only by the fresh feed flow setpoint and not by the choice of
the level setpoints, which only has a dynamic effect. We therefore distinguish between the
steady state operating degrees of freedom and the control degrees of freedom. The steady state
operating degrees of freedom is the number of independent adjustments (to valve positions or
regulatory setpoints) that affect the process steady state. For the simple example process, the

Figure 9.2. Three-tank-in-series process

steady state operating dof is 1, corresponding to the steady flow through the process, while the
control dof is 4 corresponding to the number of independent control valves. Notice that the
number of setpoints that the operator must input to the control system is 4, the same as the
number of independent valves. Of these, the level setpoints have no steady state effect. Only the
feed flow setpoint affects the steady state.
This then leads to a very simple procedure for calculating the steady state degrees of
freedom for a process. We count the number of independent control valves and subtract the
number of non-reactive surge levels as they have no effect on the steady state solution. If the
inventory however is reactive, eg level in a liquid phase CSTR, it must not be subtracted
(discounted) as the inventory (reactor holdup) affects the reaction extent (conversion) and hence
the steady state solution. We also subtract any other variables (e.g. column pressures) that must
be kept fixed at a given value for operational reasons to obtain the steady state operating degrees
of freedom.
As an illustration, consider a simple distillation column. It has six valves (including feed).
Two valves will get used for reflux drum and bottom sump level control. One valve would get
used to control the column pressure. Usually the column pressure must be maintained at the
96
design value so that temperature inferential control can be applied. Also the column feed is not in
our hands and is specified by an upstream process. Thus for a given feed and column pressure,
the steady state operating dof of a simple distillation column is 6 - 2 levels - 1 column pressure -
1 column feed = 2. The operator is free to make 2 independent adjustments. These 2 independent
adjustments may be made for maintaining 2 variables such the light key impurity in the bottoms
and the heavy key impurity in the distillate.
In Figure 9.3, we show typical steady state dofs for simple unit operations with the
implicit assumption that the feed to the unit is given (eg set by an upstream process). Figure 9.4
shows the steady state dof calculation for two example chemical processes. Notice the ease with
which dofs can be calculated without having to worry about number of independent variables
and number of independent constraints, counting which can befuddle even experienced
engineers.

9.3. Degrees of Freedom, Controller Variables (CVs) and Control Structures

The steady state operating dofs are the number of independent adjustments an operator
can make to a process that would affect the steady state solution of the process. Consider a
simple distillation column. Given the column pressure and feed rate, the operator may choose to
keep two appropriately chosen variables constant, corresponding to the two steady state dofs.
The simplest option is to fix the reflux rate(L) and the boilup(V). This is equivalent to choosing
L and V as the two column specifications. For changes in the feed rate / composition, the light
key and heavy key impurity in respectively, the bottoms and the distillate, would show
unacceptably large variation. To prevent excessive heavy key leakage down the bottoms, the
operator may choose to adjust the boilup to maintain a sensitive stripping tray temperature (TS).
To ensure that the light key leakage up the top is regulated, at least for changes in the feed flow,
the operator may choose to maintain L in ratio with the column feed F. This is equivalent to TS
and L/F as the two column dof specifications. We may similarly have the operator maintaining
TS and TR, a sensitive rectifying tray temperature, or alternatively the distillate heavy key mol
fraction (xhkD) and the bottoms light key mol fraction (xlkB). Many other choices can be made for
the 2 specification variables for simple distillation column. This example shows that there are
several options for choosing the specification variable corresponding to steady state dofs.
From the discussion above, it is apparent that holding a particular variable constant
implicitly assumes a control loop that manipulates an appropriate valve (or setpoint) in order to
maintain the variable. Figure 9.5 shows example control structures corresponding to L-V, L/F-
TS, TR-TS and xhkD-xlkB as the specification (controlled) variables on a simple distillation column.
In these structures a basic regulatory control structure is assumed where feed flow is controlled
by the feed valve, column pressure is controlled by the condenser duty and the reflux drum and
bottoms levels are controlled using respectively the distillate and bottoms.
Implicit in the pairings implemented in the structures shown in the Figure are some
common sense principles. For fast level and pressure control, the manipulated variables are
chosen 'local' to the concerned unit. Similarly, reflux is used to control a variable related to the
rectifying section (TR or xhkD) and boilup is used to control a variable related to the stripping
section (TS or xlkB). This pairing philosophy reflects the heuristic:
"Choose close by manipulated variables for controlling a process variable for a fast dynamic
pairing".

97
Figure 9.3. Illustration of control and steady state dofs for some typical unit operation

98
Figure 9.4. Illustration of dof analysis for complete chemical plants
99
Figure 9.5. Alternative CVs corresponding to steady state dofs on a simple distillation column.
(a) L-V (b) L/F-TS (c) TR-TS (d) xhkD – xlkB

100
If we let go of the "close-by" pairing philosophy, for the same set of CVs, several alternative
pairings can be proposed. Giving up close-by pairing on a unit would usually be due to plantwide
control considerations that require tighter control of a particular plant subsection. For example,
let us say the distillate from the column feeds the reaction section of a plant, where a highly
exothermic reaction occurs. We would like to hold the flow to the reactor section constant to
prevent propagation of transients to this section as it is hard to stabilize and variability in the
reaction section upsets the downstream product separation section. So now, the distillate must be
flow controlled to eliminate flow transients to the reaction section. This flow setpoint then sets
the flow through the column, instead of the column feed. Since distillate is fixed, reflux drum
level gets controlled using the reflux. The bottoms level is controlled as before using the
bottoms. Since it is important to have tight impurity control in the distillate (which feeds a
reactor), we use boil-up to control a rectifying tray temperature, as a change in boilup has an
almost immediate effect on tray temperature, unlike reflux which has a slower effect particularly
if the control tray is further down from the top. This pairing would give tighter distillate impurity
regulation. The stripping tray temperature then gets controlled using the column feed. Figure 9.6
shows four alternative pairings for TR-TS as the CVs on a column. These structures differ
particularly in the location where the flow through the column, also referred to as the throughput,
is set. Which structure should get implemented would depend on the specific plantwide context.
Even as we have not said much about plantwide control considerations, the point of the whole
exercise is to show that even for a simple distillation column with 2 steady state dofs, there exists
tremendous flexibility in the control structure that can be implemented on it due to the choice of
the specification variable corresponding to the steady state dofs as well as the pairings for the
CVs (including regulatory level and pressure loops).
How do we go about systematically choosing the CVs and the corresponding pairings is
like piecing a puzzle together. In what follows, we look at different ways of piecing together this
puzzle. The first step, as evident in what has already been discussed previously, is to count the
number of control and steady state degrees of freedom. The next step is to tabulate the different
control objectives and appropriate controlled variables (CVs) for those objectives. All control
objectives regulate some process inventory, inventory being interpreted in its most general sense
to include total material, phase, component and energy contained in a process unit and the
overall process. The regulatory control system is required to ensure (In – Out + Generation) of
the inventories in a unit and the overall process is zero so that accumulation is forced to zero to
ensure unit specific / plantwide drifts are avoided / mitigated.
The number of CVs are the same as the number of control degrees of freedom and would
encompass all inventory regulation objectives. Of these, pure surge capacities have no steady
state impact and are therefore economically not relevant. The level of component inventories in
recycle loops and product / purge streams on the other hand usually impact the steady state plant
economics significantly. The reactor operating conditions (temperature and composition) also are
usually important as the single-pass conversion and selectivity determine the cost for recycling
unreacted reactants and side-product processing cost.

101
Figure 9.6. Alternative pairings (structure) for holding TR-TS as the two steady state dof
CVS as a simple distillation column
102
9.4. Control Objectives and Choice of CVs

Given a set of control objectives and corresponding CVs plus the prioritization of the
CVs, it is relatively straightforward to devise the control loop pairings. How does one go about
systematically determining the control objectives and corresponding CVs. To the experienced
engineer, control objectives and corresponding CVs for a process are usually evident. To the
novice however, this is usually not very clear. In the following we attempt to provide a basic
framework to help figure out the control objectives and appropriate CVs.
The control system on a continuous chemical process with material and energy
integration may be viewed as an automatic mechanism for ensuring that all process inventories
are regulated at safe / optimal levels and not allowed to drift, regardless of process disturbances
such as changes in the process throughput, ambient conditions, equipment characteristics etc. All
the CVs directly/indirectly reflect process inventories; e.g. level reflects liquid inventory,
pressure reflects gas/vapor inventory, temperature reflects energy inventory and composition
reflects component inventory (inferential measurements such as column tray temperature or a
recycle flow or an appropriate separator level also indirectly reflect component inventory). Since
inventories are prone to large drifts (accumulation/depletion) unless regulated, the plantwide
control system attempts to maintain them at desired values for economic reasons or at the very
least, within an acceptable band (e.g. surge drum levels) to avoid unsafe operating conditions.
From the economic standpoint, typically component inventory levels in recycle and
product/discharge streams have a large impact on the steady state operating profit so that these
should be controlled tightly. On the other hand, surge drum levels that are part of the material
balance control system have no effect on the process steady state.
As a starting point, let us take a liquid tank with a liquid stream in and a liquid stream out
as a very simple example. If both the inlet and outlet control valves are flow controlled as shown
in Figure 9.7(a), the control structure is fundamentally flawed as it violates the overall material
balance constraint. Two flows are being independently set and any mismatch in the setpoints
would necessarily imply the liquid inventory in the tank (indicated by a level sensor) either
builds up (inflow > outflow) or depletes (inflow < outflow). The tank is then guaranteed to run
dry or over flow. In other words the implemented control system is guaranteed to fail.
The novice may argue that to satisfy the material balance constraint, both the setpoints
can be set equal. That still does not solve the basic problem as a mismatch in the two tank flows
would any way occur since sensors are never 100% accurate, the slightest of biases implying a
slow build-up / depletion in the tank level. The basic issue is that the liquid inventory in the tank
is non-self regulatory and must therefore be regulated. We need to measure (or estimate) the
liquid inventory and adjust one of the flows to ensure that the inventory is maintained within an
acceptable band. The other flow is set independently by the operator or an upstream /
downstream process. A direct measure of the liquid inventory inside the tank is its level. Figure
9.7(b-c) shows two workable control configurations that respect the material balance constraint
by controlling the tank level.
Even as the above is a very trivial example, treating a complex process with several units
and recycles as a tank and questioning if the implemented control system ensures all process
inventories (material, phase, component or energy) on each of the individual units as well as the
overall process are regulated and do not drift would reveal if the control system is workable or

103
not. We note that routine level, pressure, temperature and flow measurements that indicate
appropriate inventory levels are usually self evident.
The control structures on individual unit operations that have already been discussed in
previous chapters may be interpreted as regulating inventories. For example, in dual ended
temperature inferential LV control structure of a simple distillation column, the condenser duty
regulates the column pressure (total vapor inventory), the distillate flow regulates the reflux
drum level (reflux drum liquid inventory), the bottoms flow regulates the sump level (sump
liquid inventory), the reflux rate is adjusted to maintain a sensitive rectification section
temperature to regulate the heavy key leakage in the distillate (component inventory) and the
boilup is adjusted to maintain a sensitive stripping tray temperature to regulate the light key
leakage down the bottoms (component inventory). Each control loop on the column fixes
(regulates) a process inventory. Of these, while the two levels have no economic significance,
the light key and heavy impurity leakage levels significantly affect the column energy
consumption and are therefore economically important. The interpretation can be easily extended
to control structures on other unit operations studied earlier.

Figure 9.7. Material balance control on a liquid surge drum

(a) Unacceptable control
104 structure
(b) & (c) Acceptable control structure
9.5. Illustration of Control Objectives and CVs for Example Processes

We are now ready to illustrate control objectives and corresponding CVs for a complete
plant. Let us consider the process flowsheet in Figure 9.4(a). It has 9 control dofs and these
valves can be used for regulating 9 objectives. On the reactor, the total material hold-up and
energy hold-up must be regulated. The reactor level and temperature are appropriate CVs for the
same (1st – 2nd CVs). On the distillation column, the liquid holdup in the reflux drum and bottom
sump must be regulated. Also, the vapor hold-up in the column must be regulated. The reflux
drum and sump levels along with the column pressure are appropriate CVs for these inventories
(3rd-5th CVs). We also need to regulate the product C leakage up the top and the B impurity
leakage down the bottoms. A sensitive stripping tray temperature is a good inferential measure of
the latter (6th CV). Holding the reflux in ratio with the column feed would provide loose but
adequate regulation of the C leakage in the recycle stream (7th CV).
The remaining 2 control objectives are more subtle. By the design of the process, the
recycle stream would contain significant amounts of both the reactants, A and B, with small
amounts of C. If we look at the overall material balance across the entire plant, 1 mol A would
react with exactly 1 mol of B. The slightest excess of fresh A (or fresh B) is not allowed to leak
in the product stream due to a stringent product purity constraint and must necessarily
accumulate in the recycle loop. Unless the fresh feeds are balanced exactly as dictated by the
reaction stoichiometry, the recycle loop would slowly but surely get filled up with the excess
reactant (A or B). The recycle rate and its excess reactant composition would then increase. This
slow drift of component inventories inside the recycle loop is referred to as the snowball effect.
We need to regulate the component inventory of both the reactants in the recycle loop to ensure
stoichiometric feed balancing. This would ensure the recycle rate and its composition does not
drift. Since the reactor is inside the recycle loop, one may hold composition of a reactant (usually
the limiting reactant) to regulate its inventory (8th CV) and the total flow to the reactor to
regulate the inventory of the other component (9th CV). Note that the reactor temperature and
composition indirectly sets the production rate inside the reactor through the kinetics. We may
change either of these to bring about a change in process production rate.
As another illustration of control objectives, consider the process in Figure 9.4(b). The
process control dof is 14. The reactor material and energy inventories are reflected by reactor
level and temperature (1st – 2nd CVs). On the first column, the liquid and vapor inventories are
reflected by the reflux drum and sump levels and column pressure (3rd - 5th CVs). The column
prevents C (heavy key) leakage up the top and A (light key) leakage down the bottoms. Any A
that leaks down the bottoms would necessarily end up in the product C stream. It must therefore
be tightly regulated and a sensitive stripping section tray temperature is a good inferential
measure of the same (6th CV). Since the first column distillate is a recycle stream, loose
regulation of the C impurity in it is acceptable. Holding the column reflux to feed ratio (L1/F1)
constant should suffice (7th CV). On the second column, we again have the reflux drum / bottom
sump levels and pressure as measures of liquid and vapor inventories (8th – 10th CVs). The
column prevents B (heavy key) leakage up the top and C (light key) leakage down the bottoms.
Tight regulation of the B impurity in the product stream (component inventory) is desirable and a
sensitive rectifying tray temperature is a good inferential measure of the same (11th CV). Since
the bottoms is a recycle stream, loose regulation of the C impurity in it is acceptable. Assuming

105
that boilup is paired for tight control of rectifying tray temperature for tight product quality
control, we may hold the reflux-to-feed ratio (L2/B1) to indirectly achieve the same (12th CV).
We now consider the stoichiometric balancing of the two fresh feeds to the process. By
the design of the process, if an excess of fresh A (fresh B) is being fed, it would accumulate in
the A (B) recycle stream. The total (fresh + recycle) A (B) rate would then increase. This total
rate to the reactor then indirectly reflects the A (B) inventory in the process. We may then choose
the total (recycle + fresh) A to the reactor and total (recycle + fresh) B to the reactor as very
convenient measures of the component inventories in the recycle loops (13th – 14th CVs). As in
the previous example, the total rate of either reactant to the reactor or the reactor temperature
may be adjusted to bring about a change in the process production rate.
Table 9.1 summarizes the regulatory control objectives and corresponding CVs for the
two example processes. The relationship of the control objectives with ensuring unit specific and
plantwide material and energy balances are evident in the objectives. Comments are also
provided to highlight their economic / regulatory significance.

9.6. Snowball Effect

From the discussion above, it is evident that while the inventories that require regulation
on a specific unit are quite self-evident, figuring out recycle component inventories that require
regulation is subtler and requires some thought with respect to guaranteeing that the overall
material balance around the plant for all the components is satisfied. Material recycle introduces
high non-linearity into the process with the recycle rates being highly sensitive to small changes
in the fresh feed flow(s). This is referred to as the snowball effect.
If we consider the example process in Figure 9.4(a), its steady state dof is 6. The reactor
level and temperature and the light key / heavy key leakage in the bottoms / distillate of the
column specify four of these dofs. Let us say that we arbitrarily choose the two fresh feed rates
as specifications for the remaining 2 steady state dofs. If we try and converge the flowsheet using
a commercial simulator, we will find that if the two fresh feeds are specified to be even slightly
different, the recycle tear does not converge and keeps on blowing up. This is because the
reaction stoichiometry and nearly pure product constraint implies the reactant fed in slight excess
has no way out of the process and therefore must necessarily build up in the recycle loop. The
sensitivity of the recycle to even the slightest of mismatch between the two fresh feeds is then
infinity. If we purge a very small fraction of the recycle stream, the sensitivity of the recycle
stream rate to small changes in the fresh feed rates would still be very high, though not infinity.
This is the snowball effect.
The choice of the specification variables for the two dofs is not appropriate as the two
flows are related by overall process material balance. For robust convergence, a better
specification is specifying the total flow rate to the reactor and its A (or B) mol fraction. Both the
fresh feeds then get calculated to satisfy these two specifications.
From the operations perspective, if the fresh feed(s) are specified (ie flow controlled), the
high sensitivity of the recycle rates to the fresh feeds would cause large swings in the recycle
streams and all the equipment in the recycle loop would be subjected to large plantwide
transients for small changes in the fresh feed(s). To avoid these large swings, it is better to hold
appropriate component inventories in the recycle loop by manipulating the fresh feed(s). The
fresh feed(s) are then fed as make-up streams and only as much is fed as gets consumed. Since

106
the reactor is always inside the material recycle loop, a common industrial practice is to hold the
total reactant component feed (fresh + recycle) to the reactor constant by adjusting the
corresponding fresh feed. In cases where the recycle stream is nearly pure reactant, the
corresponding fresh feed may be adjusted to hold the total (recycle + fresh) flow constant. In
cases where the recycle stream is a mixture of reactants, appropriate composition(s) inside the
reactor and total flow to the reactor are held constant by manipulating the fresh feeds.
The basic idea of feeding fresh feeds to hold appropriate reactor conditions constant
achieves two objectives. It ensures the component inventories in the recycle loops are properly
managed. Also, by maintaining the reactor operating conditions (flow and composition) constant,
robust stabilization of the most non-linear unit operation in the process is achieved mitigating the
transients propagated to the downstream separation section.

107
 Table 9.1. Regulatory objectives and CVs for the two example processes
SNo Regulatory objective CV Significance
Single column recycle process
Closes reactor MB*. Affects
1 Reactor liquid inventory Reactor level
conversion and separation load.
Closes reactor EB**. Affects
2 Reactor energy inventory Reactor temperature
conversion and separation load.
Column reflux drum liquid
3 Reflux drum level Closes reflux drum MB.
inventory
4 Column sump liquid inventory Sump level Closes sump MB.
5 Column vapor inventory Column pressure Closes column EB.
Closes the lk/hk balance on the
column. Affects column steam
6 Distillate hk& (C) leakage Reflux to feed ratio consumption. Bottoms B leakage
7 Bottoms lk% (B) leakage Stripping tray temp fixed by min product quality. Too
much distillate C leakage dilutes
reactor reducing conversion.
8 Component B circulating in recycle Reactor B mol fraction Fixes recycle stream conditions
and hence affects column steam
9 Component A circulating in recycle Total feed to reactor consumption.
Two-column recycle process
Closes reactor MB and EB.
Reactor level and
1-2 Reactor liquid and energy inventory Affects conversion and
temperature
downstream separation load.
Column reflux drum
3-6 Reflux drum/sump liquid inventories Closes reflux drum/sump MBs
and sump levels
7-8 Vapor inventory in columns Column pressures Closes column EBs
Closes the lk/hk balance on the
Reflux to col feed columns. Affects reboiler steam
9 Column 1 distillate C (hk) leakage ratio consumption. Too much C leakage
10 Column 2 bottoms C (lk) leakage Reflux to col feed in recycle streams dilute reactor
11 Column 1 bottoms A (hk) leakage ratio reducing conversion. Col1 bottoms
12 Column 2 distillate B (hk) leakage Stripping tray temp A leakage and Col2 distillate B
Rectifying tray temp leakage set by max product
impurity specification

13 Component A circulating in plant Total# A to reactor Fixes recycle stream conditions

and hence affects column steam
14 Component B circulating in plant Total# B to reactor consumption.
*: material balance; **: energy balance; &: heavy key; %: light key; #: recycle plus fresh feed

108
 Chapter 10. The Pairing Issue: Selection of MVs for CVs
Given a set of inventory regulation control objectives and corresponding CVs, the next
step is to select the manipulated variable (MV) pairing for each of the CVs. To select pairings for
the CVs, they must be prioritized with the pairing for the highest priority CV being selected first
followed by the pairing for next one and so on so forth. Different prioritizations would lead to
different pairings and hence different control structures.

10.1. Conventional Pairing Approach

The conventional approach to designing the loop pairings is to first choose the process
variable that is adjusted for setting the throughput. The setpoint corresponding to that process
variable control loop is referred to as the throughput manipulator (TPM). Conventionally, the
throughput manipulator is chosen at a fresh feed to the process. Other TPM locations are possible
and include the product stream flow for on-demand process operation, where the demand from a
customer must be immediately met; an intermediate process stream flow for mitigating transients
to the connected unit; directly setting reactor temperature or limiting reactant concentration in a
process with a reactor etc.
With the TPM in place, local inventory loops on each of the units are then put in place to
establish total material balance / energy balance control. By local, we mean that the MV for
controlling the inventory is local to the unit containing the inventory. This is illustrated in Figure
10.1 for the 'three tanks in series process', where the throughput may be set at any of the four
process streams. The tank level controllers upstream of the TPM (set flow) are then naturally
oriented opposite to the process flow while the level controllers downstream of the TPM are
oriented in the direction of process flow. The upstream level controllers act to supply the set flow
while the downstream level controlled act to process the set flow. The total material balance
control structure thus radiates outwards from the TPM. Local loops for energy balance control
would usually include temperature control of an exothermic reactor using reactor cooling duty
stabilizing the most non-linear unit in the plant.
With the basic material balance / energy balance control pairings in place, the pairings for
the remaining CVs are chosen from the remaining valves. These involve loops for regulating
component inventories and are usually economically important. In cases where the open loop
response of the CV is sluggish, an appropriate cascade arrangement is implemented with a slave
controller holding a faster secondary variable and the master controller holding the primary
variable by adjusting the slave loop's setpoint.

10.2. Luyben's Pairing Approach

In the first significant departure from the conventional pairing approach, Luyben et al.14
insightfully noted that since non-reactive surge inventories have no steady state economic
impact, material balance control loops should have lower prioritization so that the best pairings
get implemented for the tightest control of economically important CVs. Their prioritization
hierarchy thus first fixes the TPM and energy balance control, then establishes loops for
economically important objectives (quality, safety, effluent discharge etc) and finally pairs loops
for material balance (material inventory) control.
109
 Figure 10.1. Alternative TPM locations and material balance control in three tanks-in-series
process

10.3. Regulatory Plantwide Control Structure Synthesis Examples: Conventional vs

Luyben’s Approach
We are now ready to synthesize and contrast plantwide control structures using the
conventional approach and Luyben’s approach. For continuity, we consider the two example
processes in Figure 9.4.

10.3.1. Single Column Recycle Process

In the conventional approach, the TPM is chosen at a process fresh feed. Let us say the
fresh B feed (FB) is the TPM (1st loop). The reactor temperature (Trxr) is then controlled using its
cooling duty (Qrxr), which would provide tight temperature control to regulate the reactor energy
balance (2nd loop). Its level (LVLrxr) is controlled using the total flow out of the reactor (F1) to
110
the column (3rd loop). On the column, the pressure (Pcol) is controlled using the condenser duty
(Qcnd), the reflux drum level (LVLRD) is controlled using the distillate (D1) and the sump level is
controlled using the bottoms (B1) (4th – 6th loops). The impurity B mol fraction in the product
stream (xBB1) is regulated in a cascade arrangement by adjusting the setpoint of a sensitive
stripping tray temperature (TScol) which manipulates the column boilup (V1) (7th loop). The C
impurity in the distillate (xCD1) is loosely regulated by holding the reflux in ratio with the column
feed (L1/F1) (8th loop). Lastly, the B mol fraction in the reactor (xBrxr) is maintained by adjusting
the fresh A to fresh B ratio setpoint (9th loop). Maintaining fresh A in ratio with fresh B ensures
the two fresh feeds move together in (near) stoichiometric ratio and large imbalances in the
reactant feeds are avoided. The conventional control structure is shown in Figure 10.2(a). Note

Figure 10.2(a). Conventional control structure with TPM at fresh B

that since FA is flow controlled, large transient swings in the recycle rate due to the snowball
effect are likely with the recycle rate floating to the appropriate value.
In Luyben’s approach for plantwide control structure design, the exothermic reactor
energy balance regulation loop is first implemented so that a potential instability is first
stabilized. The conventional Trxr-Qhtr pairing is implemented for tight energy balance regulation
(1st loop). We assume the TPM can be placed anywhere in the process and there is no operational
111
constraint such as on-demand operation or a process feed set by an upstream process. Where to
locate the TPM is then left as a decision to be taken later. The next loop to be implemented then
is the product purity control loop. For tight regulation of xBB1, a cascade arrangement is
implemented with the xBB1 adjusting the setpoint of the TScol controller which manipulates the
column boilup (V1) (2nd loop). In the absence of any other information, the next loops to be
implemented are ones for feeding the fresh feeds as make-up streams. The total flow to the
reactor (Frxr) is maintained by adjusting FB (3rd loop). FA is maintained in ratio with FB and its
setpoint is adjusted to maintain xBrxr (4th loop). With these two loops, the recycle rate and
composition are not allowed to float or float only within a very narrow band. Snowballing is thus
mitigated. We are now ready to put in the material balance control system. The pairings LVLrxr-
Fcol, LVLRD-D1, LVLbot-B1 and Pcol-Qcnd are chosen for regulating the liquid and vapor
inventories in the process (5th – 8th loops). Lastly, the L1/F1 ratio loop is chosen for managing the
column reflux (9th loop). The control structure obtained is shown in Figure 10.2(b). Even as it
‘looks’ very similar to the conventional structure (Figure 10.2a), the design philosophy including
how fresh feeds are managed and the prioritization of the control objectives is very different. To
manipulate the throughput, we may adjust either of the Trxr, xBrxr or Frxr setpoints. Usually Trxr is
not adjusted as the catalyst has a very narrow operating temperature range for which the
manufacturer guarantees catalyst life. Also, usually the reactor must be operated with one of the
reactants being limiting which would fix xBrxr. FrxrSP is then the only option for the TPM.

Figure 10.2(b). Luyben’s control structure with TPM at reactor inlet

112
10.3.2. Two Column Recycle Process
The conventional plantwide control structure for the two column recycle process (Figure
9.4b) is synthesized as follows. Let us say the fresh B (FB) is the TPM (1st loop). The reactor
temperature (Trxr) is controlled using the reactor cooling duty (Qrxr) for tight energy balance
regulation on the most non-linear process unit (2nd loop). Material balance control consists of
controlling reactor level (LVLrxr) using reactor outlet flow (Fcol1), the two reflux drum levels
(LVLRD1 and LVLRD2) using the respective distillate flows (D1 and D2), the two column sump
levels (LVLbot1 and LVLbot2) using the respective bottoms flows (B1 and B2) and the two column
pressures (Pcol1 and Pcol2) using the respective condenser duty valves (Qcnd1 and Qcnd2) (3rd to 9th
loops). We now implement component inventory control loops. On the first column, the reflux is
maintained in ratio with the feed to provide loose regulation of the C impurity in the A recycle
stream (10th loop). A sensitive stripping tray temperature (TScol1) is maintained by adjusting the
boilup (V1). The temperature setpoint is adjusted by an A impurity in product (xAD2) controller in
a cascade arrangement (11th loop). On the second column, the reflux is maintained in ratio with
the feed and the L2/B1 ratio setpoint is adjusted by a B impurity in product (xBD2) controller (12th
loop). The column boilup (V2) is manipulated to hold a sensitive stripping tray temperature
(TScol2) constant to regulate the C leakage down the bottoms (13th loop). The last loop must
ensure that FA exactly balances FB (TPM) to satisfy the overall plant material balance through the
reaction stoichiometry. The total (fresh + recycle) A rate (FTotA) to the reactor is maintained by
adjusting FA (14th loop). The control structure is shown in Figure 10.3(a). Note that in this
control scheme, the B recycle can show large swings due to the snowball effect.
We now synthesize the regulatory plantwide control structure using Luyben’s pairing
approach. The Trxr-Qrxr pairing is first selected for robust stabilization of the reactor energy
balance (1st loop). As in the previous example, we assume that the TPM can be chosen anywhere
in the plant and leave the decision for later. The next loops to be implemented are for tight
product impurity control. The two impurities in the product are A leaking down the first column
and B leaking up the second column. For tight regulation of the former, the TScol1-V1 pairing is
selected with the temperature setpoint cascaded by a xAD2 controller (2nd loop). For tight
regulation of xBD2, a sensitive rectifying tray temperature in the second column (TRcol2) is
maintained by manipulating V2 with its setpoint cascaded by the xBD2 controller (3rd loop). Tray
temperature control using boilup achieves the tightest temperature control on a column. Here,
this dynamic advantage of the pairing is leveraged for achieving tighter B impurity control than
the conventional pairing with reflux rate (or ratio). With the product impurity loops in place, we
implement loops for feeding the fresh feeds as make-up streams. The total (fresh + recycle) B
(FTotB) to the reactor is maintained constant by manipulating FB (4th loop). The total (fresh +
recycle) A (FTotA) to the reactor is maintained by adjusting FB and its setpoint is maintained in
ratio with FTotB (5th loop). Maintaining FTotA and FTotB using the fresh feeds ensures the
unreacted A and B component inventories in the recycle loops are tightly regulated to mitigate
snowballing. Maintaining FTotA in ratio with FTotB mitigates the transient variability in the reactor
composition. The pairings LVLrxr-F1, LVLRD1-D1, LVLRD2-D2, LVLbot1-B1, LVLbot2-B2, Pcol1-
Qcnd1 and Pcol2-Qcnd2 are implemented to control the process liquid and vapor inventories (6th –
12th loops). The last two loops to be implemented are holding the two column reflux rates in ratio
with the column feeds (L1/F1 and L2/B1) (13th – 14th loops). In conjunction with the temperature
loops on the two columns, these two loops ensure the impurity leakage in the two recycle
streams is loosely regulated. The control structure is shown in Figure 10.3(b). To manipulate the

113
throughput, Trxr, FTotA or FTotB may be adjusted. Usually, one is not free to adjust Trxr. Also, the
reactor must be operated with a minimum excess of one of the reactants (say A). The total
limiting reactant (B) flow to the reactor (FTotB) would then be an appropriate TPM. We again
highlight that even as the structures in Figure 10.3(a-b) ‘look’ similar, their synthesis
philosophies are very different.

114
Figure 10.3(a). Conventional control structure for two column recycle process

115
Figure 10.3(b). Luyebn’s control structure for two column recycle process

116
 Chapter 11: Economic Considerations in Plantwide Control
Given a regulatory plantwide control structure that ensures the unit specific and overall
material and energy balances are satisfied so that the process inventories do not drift or drift
within an acceptably small band, we are ready to bring in economic considerations. The key
question is, “What are the process inventories that significantly affect steady operating profit and
their optimal levels (values)?” Engineering common sense applied to a process would usually
reveal the economically important inventories and we discuss some of the considerations below.

11.1. Economic Process Operation Considerations

From the economic point standpoint, on-aim product purity is always desired. The
product then contains maximum allowed impurity for zero product give-away or alternatively,
for selling maximum allowable cheap impurities for the price of the product (legal adulteration!).
Because process raw materials (reactants) are usually quite expensive (much much more than
energy), their loss in non-product streams (eg a purge stream or a waste-product stream)
discharged from the process must be regulated tightly at an acceptably small value. This includes
minimizing the loss of expensive reactants as undesired by-products that are discharged from the
plant, since the waste product consumes expensive reactants with no sales revenue.
In reactors, there usually exists a single-pass conversion versus selectivity (yield to
desired product) trade-off. Side reactions always occur in any reactor and these are often
suppressed by designing the reactor to operate in large excess of a reactant. One would like to
maximize the single-pass reactor conversion to reduce the amount of unreacted reactants to be
recycled and hence the associated recycle cost. For irreversible reactions, this would correspond
to operating the reactor at the maximum allowed temperature. However, because the activation
energy of the side reaction(s) is higher than the main reaction with the catalyst significantly
reducing the activation barrier for the main reaction, the %age increase in reaction rate per unit
temperature increase is higher for the side reaction. Thus for irreversible catalytic reactions, any
increase in conversion via an increase in temperature comes at the expense of reduced yield to
desired product. The reactor temperature is then likely to have an optimum conversion-yield
trade-off with higher single pass conversion reducing the recycle cost (lower unreacted reactants
to be recycled) at the expense of lower yield to desired product. If the process is such that the by-
product is simply discharged from the process, the loss in yield dominates since energy is
significantly cheaper than raw materials and the reactor operating conditions must be chosen to
maximize yield. This would usually correspond to maximizing the excess reactant composition
in the reactor, usually limited by a recycle equipment capacity constraint, along with an optimal
temperature for high yield (say >95%) and not-too-low a conversion. In cases where the by-
product is further processed back to the desired product, there is an associated processing cost
which goes up as the by-product formation rate goes up (with increase in temperature). The
reactor temperature would then still have an optimum; however since both reactant recycle cost
and side-product processing cost primarily correspond to energy consumption (which is cheap),
it would usually be optimal to have lower than maximum achievable excess reactant in the
reactor and a higher operating temperature (as no by-product is discharged).
Unlike the reactor temperature, the reactor hold-up (level for liquid phase reactors and
pressure for gas phase reactors) affects all the reaction rates equally with a eg 10% increase in
117
hold-up causing a 10% increase in all reaction rates. For kinetically limited reactors (ie all
irreversible reactions and reversible reactions where the reactor is not large enough for
equilibrium to be attained), it is then always optimal to operate at maximum reactor hold up
(maximum level for liquid phase CSTRs and maximum pressure for gas phase reactors) as we
get an increase in conversion with no yield penalty.
For optimal operation, the total energy consumption per kg product should generally be
as small as possible. Heuristics for energy efficient operation of common unit operations are
well-known and should be liberally applied. This includes preventing over-refluxing in
distillation columns by dual-ended control, efficient operation of furnaces by adjusting the fuel to
air ratio to maintain stack-gas composition, floating pressure control of a superfractionator, using
valve position control on a variable speed pump feeding parallel process trains etc. These
heuristics have been discussed earlier.

11.2. Process Operation Modes

Continuous chemical processes are usually operated in 2 modes. In Mode I, the process
throughput (production rate) is specified based on market demand-supply considerations and
economic operation is equivalent to maximizing process efficiency (eg minimum steam
consumption per kg product or maximum yield to desired product etc). In Mode II, the market
conditions are such that it is optimal to operate the process at maximum (economic) throughput.
Plants immediately after commissioning are often operated at maximum throughput to maximize
revenue and pay-off debts. First-to-patent product / process monopolies may also be operated at
maximum throughput given sufficient product demand.

11.3. Process Constraints and Economic Operation

The discussion on economic considerations hints at economic process operation requiring

operation at or close to constraints. The constraints may be soft, where short duration constraint
violations are acceptable, or hard, where constraint violations are unacceptable or not possible.
Process operation at the maximum allowed product impurity constraint for no product give-away
is an example of a soft constraint. Hard constraints usually correspond to equipment capacity
constraints. Examples include operating a gas recycle compressor at maximum duty to maximize
gas recycle rate and hence minimize fresh gas consumption, operating a distillation column at its
flooding limit (maximum boilup) to maximize the recycle of the excess reactant for suppressing
a side reaction etc.
At the design throughput, hard equipment capacity constraints are usually not active (due
to equipment overdesign). However, as throughput is increased, equipment successively hit
capacity constraints. For example, the boilup in a distillation column is commonly manipulated
for stripping tray temperature control. As throughput is increased sufficiently, the boilup would
increase to a point where the column approaches its flooding limit with the high boilup not
allowing liquid to drop down the trays. Upon hitting the flooding limit (maximum boilup, VMAX),
tray temperature control would be lost. The loss in tray temperature control would imply loss in
regulation of the light key dropping down the column. Let us say the column bottoms stream is a
product stream. Product light key impurity control is then lost, which is unacceptable. If the
bottoms is a recycle stream, the light key inventory in the recycle stream is unregulated and can

118
build-up (snowballing) unless the throughput is cut. The point is that as constraints go active,
regulation of crucial control tasks may be lost.

11.4. Approaches for Handling Equipment Capacity Constraints

11.4.1. Backed-off Operation

How does one handle equipment capacity constraints going active? Consider the simple
distillation column with conventional single ended temperature control using boilup and
maximum boilup (VMAX) representing a capacity constraint. The simplest thing to do would be to
back-off the column feed sufficiently so that VMAX does not go active for the worst expected
disturbance. This is illustrated in Figure 11.1(a). The maximum achievable steady throughput
would then be lower, representing an economic loss.

11.4.2. Use of Valve Positioning (Optimizing) Controller

To automate the back-off in throughput, one may implement a valve positioning

controller that maintains the boilup at a specified value by manipulating the feed rate. This is
shown in Figure 11.1(b). Since an adjustment in feed by the VPC would affect the boilup
reasonably quickly through the action of the temperature controller, the back-off would be lower
than what was necessary using the strategy in Figure 11.1(a). Even so, some back-off would be
necessary representing a loss in maximum throughput ie an economic loss.
In the control system in Figure 11.1(b), the VPC setpoint sets the feed to the column and
thus indirectly acts as the TPM. A simple and effective control scheme for handling the VMAX
constraint is to directly use the boil-up flow setpoint as the TPM and control tray temperature
using the column feed, as shown in Figure 11.1(c). Increasing the boilup would cause the tray
temperature to increase and the temperature controller would increase the cold fresh feed to bring
the increasing temperature back to setpoint. The temperature control would be reasonably tight
as long as the control tray is not too far below the feed tray. Notice that due to tight control of the
boilup using reboiler duty, little/no back-off from the VMAX limit would be necessary so that the
process can be operated at VMAX with no (or negligible) loss in maximum achievable throughput.

11.4.3. Altering Material Balance Control Structure Using Overrides

There is also the conventional approach of handling constraints using override

controllers. The VMAX constraint on a distillation column is conventionally handled by a slower
override tray temperature controller with its setpoint slightly below the nominal setpoint and its
output passing to the column feed valve through a low select, as shown in Figure 11.2(a). When
VMAX is inactive, the nominal temperature controller controls tray temperature close to the
nominal setpoint. The tray temperature is then higher than the override temperature controller
setpoint so that its output increases in an attempt to put more cold feed to reduce the tray
temperature to its setpoint. The output is then high and the low select on the signal to the feed
valve passes the desired feed throughput signal (column feed as TPM). When the VMAX
constraint goes active on eg sufficiently increasing column feed rate, the tray temperature would

119
decrease causing a decrease in the override controller output with the low select eventually
passing feed manipulation to the override temperature (column feed under temperature control).
The override scheme thus alters the control structure from fixed feed – manipulated boilup to
fixed boilup – manipulated feed.
In case the feed to the column is being set by an upstream process eg by the level
controller of the upstream reactor, the temperature override taking up column feed manipulation
would imply loss of level control on the reactor. The reactor level would then increase and an
override level controller with its setpoint slightly higher than the nominal level controller
setpoint must now take up manipulation of reactor feed to regulate its level. Appropriate
overrides will have to be implemented all the way back to the process feed, as shown in Figure
11.2(b-c). Regardless of the number of intervening units between the process feed and the
constrained unit, what the override scheme does is alter the material balance control structure
from fixed process feed – varying constraint variable (boilup in the distillation example) to fixed
constraint variable – varying process feed.

11.4.4. Using Constraint Variable as Throughput Manipulator

The use of overrides for altering the material balance control structure on hitting a
constraint can be avoided as illustrated in Figure 11.3. Here, the constraint variable is the TPM
and the material balance control loops are oriented around it using the radiation rule. Clearly, this
gives a much simpler control system with no overrides. Also, no (minimum) back-off is needed
from the active constraint limit. In contrast, a major disadvantage of using overrides is the need
for appropriate offset in override controller setpoints. In the Figure 11.2 examples, the nominal
reactor level setpoint would necessarily be lower than maximum implying that the nominal
process operation would be at a lower than maximum single pass conversion due to lower than
maximum holdup with consequent higher recycle cost. Similarly, the offset in the column
temperature override controller would imply higher steady loss of the light-key down the
bottoms once VMAX goes active. The overrides also introduce an inherent dynamic disadvantage
with the overrides taking time to take-over and give up control and also an element of on-off
control with potential repeated misfiring causing unnecessary plantwide transients, particularly
when the final steady state is not at the constraint limit but slightly below it. In our considered
view, the use of overrides should be minimized as far as possible and using a (hard) equipment
capacity constraint variable controller setpoint as the TPM and orienting the material balance
control system around constitutes a simple and effective way of handling one such hard
constraint variable for negligible back-off and consequent economic loss.
Typically the maximum throughput solution has multiple hard active constraints. The
economic loss due to a back-off from these constraints would usually be the largest only with
respect to a particular constraint. We refer to this constraint as the economically dominant
constraint. For economic operation, we choose this constraint variable (or setpoint of the loop
that controls it) as the TPM and put in place the total material balance control system around it.
This minimizes the back-off in the economically dominant constraint mitigating the consequent
economic loss. The loss in control dofs due to the remaining hard active constraints is then
managed with sufficient back-off from the constraint limits which causes only an acceptably
small steady economic loss, since these constraints are not economically dominant.

120
Figure 11.1. Various control scheme for handling equipment capacity constraint

121
Figure 11.2. Override control scheme for handling capacity constraint

122
Figure 11.3. Choosing TPM at the constraint variable to avoid overrides

123
 Chapter 12. Economic Plantwide Control Examples
We are now ready to synthesize a plantwide control structure for economic operation of the two
example chemical processes in Figure 9.4 using the engineering heuristics discussed above.

12.1. Single Column Recycle Process

The material, component, phase and energy inventories have already been discussed
previously. We now bring in economic considerations. The process has 6 steady state dofs. Since
there are no side reactions in this toy-problem, economic operation corresponds to minimizing
energy consumption (i.e. column reboiler duty). If the separation in the column is relatively easy
(likely as C is formed by the addition of A to B and is therefore significantly heavier than both
reactants), minimizing energy consumption per kg throughput would correspond to maximizing
single pass conversion and hence minimizing the recycle load. Accordingly, the reactor should
be operated at maximum level (LVLrxrMAX) and temperature (TrxrMAX). Also, no product give-
away requires the B impurity in the product to be at its maximum allowed limit (xBB1 MAX). These
three constraints would be active regardless of throughput (ie both in Mode I and Mode II) and
account for three steady state dofs.
In Mode I, the throughput (FA) is specified leaving 2 unconstrained dofs. These
correspond to the C leakage in the recycle stream and the B composition in the reactor (xBrxr) or
more generally, in the recycle loop. If too little C leaks up the top (sharp separation), the boil-up
increases (higher reflux for the sharper rectification). On the other hand, if too much C leaks up
the top, the reactor gets diluted with the recycle C and the reactor reactant composition goes
down for lower single pass conversion and consequent higher recycle cost. Sufficient reflux thus
needs to be provided in the column so that too much C does not leak up the top. This is achieved
by maintaining the reflux in ratio with the column feed (L1/F1) ensuring adequate C regulation at
all throughputs.
With respect to xBrxr, we note that the conversion would be maximized for comparable
reactor A and B mol fractions as the irreversible reaction kinetic expression is
r = k xArxr xBrxr
Now since the reactor contains C (generated by reaction) and its amount varies with throughput
(generation rate), the optimal value of xBrxr that ensures xBrxr ≈ xArxr, would vary with throughput.
Care must then be exercised that the specified xBrxr setpoint is not infeasible due to the variation
in xCrxr. The optimum xBrxr would be the smallest at maximum production (largest xCrxr) large. To
ensure feasibility the desired setpoint over the entire throughput range, we may choose to
implement this setpoint value at all throughputs. At low throughputs (xCrxr small due to low
generation, xBrxr specified to be small), the reactor then gets operated in significant excess A
environment implying higher than necessary reboiler duty.
One way around this problem is to realize that the recycle stream contains mostly A and
B with only a small amount of C. If instead of holding xBrxr constant, we ensure that xBD1 ≈ 50%
(ie comparable A and B in recycle stream), then xBrxr would automatically float to be comparable
to xArxr. Now since B is heavier than A and therefore requires more energy to boil-off, a
reasonable specification for near optimal operation over the entire throughput range would be
holding xBD1 slightly but not too far below 50% (say at 45%). Such a choice would ensure

124
reactor operation close to maximum achievable single pass conversion (an economic objective)
across the entire throughput range.
As throughput is increased, let us say that the column approaches flooding. The
maximum boilup (VMAX) then limits the maximum achievable throughput (Mode II operation).
We take the two regulatory plantwide control structures synthesized earlier (Figure 10.2) and
adapt them for economic operation over the entire throughput range.
In Figure 12.1(a), we take the conventional plantwide control structure with FB as the
TPM and modify it for economic operation. The setpoints for Trxr and LVLrxr loops are specified
to be TrxrMAX and LVLrxrMAX (for maximum single pass conversion). A slow xBD1 controller is
implemented that adjusts the xBrxr composition loop setpoint to hold xBD1 at its (near) optimal
value (chosen as 45% here) for the entire throughput range. Similarly, L1/F1 SP is set at an
appropriate value for ensuring too much C does not leak in the recycle stream over the entire
throughput range. For maximum throughput operation with V1MAX as the bottleneck constraint,
an override scheme for altering the material balance control structure is implemented. Notice that
the setpoint of the nominal and override temperature controllers on the column comes from the
master xBB1 (product B impurity) controller. The override temperature controller setpoint is
always slightly lower than the nominal setpoint via the negative bias. When the temperature
override gets triggered, the product impurity would increase (as override temperature setpoint is
lower) and the action of the xBB1 controller would slowly bring it back to the appropriate level.
On the other hand, when the nominal controller takes up temperature control (V1MAX goes
inactive), since its setpoint is higher than the override setpoint, the impurity leakage would
decrease (below maximum allowed) and then get back to the desired value via the action of the
xBB1 controller. Clearly, product impurity control becomes loose due to the overrides ‘taking
over’ or ‘giving-up’ control.
To avoid the disadvantages associated with overrides, one may insist on having a fixed
control structure regardless of throughput. If the conventional regulatory control loops are
already implemented and are not modifiable, the only free setpoint available for maintaining the
constraint variable (V1) at a desired value is FBSP. This loop is shown in Figure 12.1(b) and is a
long one. When coupled with the snowball effect, V1 would only get controlled loosely around
the desired setpoint implying a large back-off from V1MAX and consequent throughput loss.
We may also take the regulatory control structure synthesized using Luyben’s approach
and adapt it for economic operation. Figure 12.2(a) shows the adapted control structure along
with a material balance altering override scheme for handling the V1MAX constraint for maximum
throughput operation. Figure 12.2(b) shows a long V1 constraint control loop manipulating Frxr to
avoid the use of override controllers. These modifications to the basic regulatory control
structure are very similar to those for the conventional control structure and are therefore not
elaborated upon. It is however worth mentioning that tighter V1 control by the long V1-Frxr loop
would be achieved as the snowball effect is mitigated with the fresh reactants being fed as make-
up streams. The back-off from V1MAX would then be lower and the control scheme would
achieve higher maximum throughput than the one in Figure 12.1(b).

125
 (a)

(b)
Figure 12.1. Handling capacity constraint in single column process (Conventional Process)
(a) Using overrides (b) Using long active constraint control loop
126
 (a)

(b)
Figure 12.2. Handling capacity constraint 127
in single column process (Luyben structure)
(a) Using override (b) Using long active constraint control loop
In Figure 12.3, we show the control system with V1SP as the TPM and the material balance
control loops oriented around it. For economic operation, the reactor is operated at TrxrMAX and
LVLrxrMAX. Also, a slow xBD1 controller that cascades a setpoint to the xBrxr controller is
implemented for ensuring near maximum reactor conversion at all throughputs. The control
structure is particularly elegant in terms of the simplicity with which the V1MAX active constraint
is handled with no overrides. The operator simply increases V1SP to V1MAX to transition to
maximum throughput. More importantly, unlike the other control structures, the basic material
balance control structure remains the same regardless of throughput. The only potential
disadvantage is slightly more loose product impurity control at low throughputs (where V1MAX is
inactive) as the boilup is not used for column temperature column. Appropriate detuning of other
loops, in particular the surge level loops, to mitigate the transients propagated to the column can
however be easily applied to ensure the product quality control is acceptably tight. Advanced
control algorithms may also be applied to mitigate the variability in the product quality. The
control structure is thus the simplest possible solution for economic process operation over the
entire throughput range (low to maximum throughput).

Figure 12.3 Using constraint as TPM to avoid overrides on the single column recycle process

128
12.2. Two Column Recycle Process

This process has 8 steady state dofs, as discussed earlier. Purely for the sake of a more
interesting discussion, let us assume that there is a side reaction (assume side product volatility is
such that it leaves with product C stream) and that this side reaction is suppressed by operating
the reactor in excess A environment (B limiting). Economic process operation then requires
maximizing the reactor excess A environment, which requires operating the first column at
maximum boilup (V1MAX) so that the A recycle rate is as high as possible. To maximize single-
pass conversion with no yield penalty, it should be operated at maximum level (LVLrxrMAX).
Also, the A and B impurities in the product should be at their maximum limits for no product
give-away (xAD2 MAX and xBD2 MAX). These four constraints are active at all throughputs. In Mode
I (given throughput), we have a specified throughput leaving 3 unconstrained steady state dofs.
These correspond to the optimum reactor temperature (conversion-yield trade-off) along with the
C leakage in the A recycle stream and in the B recycle stream. This C leakage must be kept small
enough at all throughputs. As throughput is increased, let us say the maximum boilup on the
second column (V2MAX) constraint is hit, which fixes the maximum achievable throughput (Mode
II).
We now adapt the conventional plantwide regulatory control structure (FB TPM) for
economic operation (Figure 10.3a). The adapted control structure is shown in Figure 12.4(a). In
the regulatory control structure, the product impurity control loops are already in place and their
setpoints are set at the maximum acceptable impurity level (xAD2 MAX and xBD2 MAX). The reactor
level setpoint is specified at LVLrxrMAX. To operate close to V1MAX, a V1 controller is
implemented which manipulates FTotA/FTotBSP in a long loop. Its setpoint will require sufficient
back-off from V1MAX to ensure A impurity regulation is never lost. The reactor temperature
setpoint is specified at an appropriate value that ensures the yield is always sufficiently high. On
the first column, L1/F1 setpoint is fixed at a value that ensures too much C does not leak up the
top over the entire throughput range. On the second column, the stripping tray temperature
setpoint is chosen to regulate C leakage down the bottoms at an acceptably small value. For
handling the bottleneck V2MAX constraint that limits maximum throughput, a material balance
altering control scheme with overrides from the second column back to the fresh A feed is
implemented. Note that V2MAX represents a capacity constraint on the amount of product C that
can be boiled off. If too much C is generated in the reactor than can be boiled off in the second
column, the extra C would necessarily accumulate in the B recycle stream. The override scheme
acts to cut the fresh B feed to the appropriate value so that the C generation in the reactor exactly
matches what is boiled off in the second column. If the override scheme for altering material
balance structure is to be avoided, FBSP must get adjusted to hold V2 (constraint variable) in a
long loop. While it may be acceptable to let the C impurity in the recycle stream float for short
durations till the long V2 loop sufficiently reduces FBSP after V2MAX goes active, large plantwide
transients due to adjustment in FB (snowball effect) are likely and conservative operators may
simply back-off V2SP sufficiently to ensure V2MAX never goes active.

129
 (a)

(b)
Figure 12.4 Use of overrides for handling capacity constraints for the two column recycle
process. (a) Conventional structure (b) Luyben’s structure

130
Figure 12.4(b) shows the adapted control structure for economic operation with regulatory
plantwide control structure from Luyben’s approach. The adaptations are very similar to the
conventional structure (Figure 12.4a). Note that the L2/B1 ratio controller must be specified to a
value that ensures too much C does leak down the second column bottoms over the entire
throughput range. To avoid the override scheme for altering material balance control when
V2MAX goes active, one can adjust FTotBSP to maintain V2 in a long loop. The plantwide transients
are expected to be smooth as FTotB is inside the recycle loop so that FB is always fed as a makeup
stream mitigating the snowball effect and the back-off from V2MAX would be smaller.
In this example, we have two hard equipment capacity constraints, V1MAX and V2MAX. In
the synthesized control structures, some back-off from V1MAX and V2MAX is needed to avoid loss
of product quality control and snowballing issues. The back-off from V1MAX causes a loss in
selectivity while and back-off from V2MAX causes throughput loss. The latter can be a significant
economic loss and to avoid the same we may use V2SP (last constraint to go active) as the TPM
and orient the material balance control system around it as shown in Figure 12.5. TScol2 is
controlled using B1, LVLbot1 is controlled using Fcol1 and LVLrxr is controlled using FTotB. As
before, FTotA is maintained in ratio with FTotB to ensure the reactor feed composition does not
vary too much. The ratio controller also ensures tight reactor level control with the total reactor
feed varying in response to a change in its level. The rest of the control system is self
explanatory.
Can we further alter the control structure to ensure the back-off from V1MAX is also
eliminated. We show one possible control structure (there are other possibilities too) in Figure
12.6. Here, V2SP is used as the TPM as before. Since V1MAX is active, it is not used for
controlling TScol1 and Fcol1 is adjusted instead to ensure the A impurity in the product is always
regulated. LVLbot1 is then controlled using B1 and LVLRD1 is controlled using D1. Similarly
LVLRD2 and LVLbot2 are regulated using D2 and B2 respectively. LVLrxr is controlled using FTotA
with FTotB maintained in ratio to ensure the proper A excess in the feed to the reactor. The
column pressures are controlled using the respective condenser duty valves. For product impurity
control, the xAD2 controller adjusts the TScol1 controller setpoint while the xBD2 controller adjusts
L2/B1, as before. On the second column, no close by valves are available for stripping tray
temperature control and the C leakage in B2 remains unregulated. V2SP (TPM) fixes the product
C boil-off from the second column and if more C is being generated in the reactor than what is
boiled-off, it would drop down the second column and B2 can show a very large increase
(snowballing). To mitigate the same, B2 is loosely regulated by adjusting the FTotB/FTotA SP. If B2
increases, the ratio setpoint is increased causing a decrease in FB with FA also eventually
decreasing so that only as much C is produced in the reactor as is being boiled off in the second
column. Loose control of B2 flow rate is acceptable as it is a recycle stream and not an exit
(product, byproduct or purge) stream. This example illustrates that economic considerations, in
particular, tight control of equipment capacity constraints, results in a plantwide control structure
that is very different from structures synthesized using the conventional approach or Luyben's
approach.
The two toy problems considered here illustrate how economic considerations impact
plantwide control structure design. We also hope that the elaborate discussion for the two case
studies convinces the readers that common sense based process engineering principles clearly
bring out the major considerations in economic / efficient process operation, at least at the
qualitative level. These economic considerations, including equipment capacity constraints,

131
translate to economic control objectives, which then govern the pairings to be implemented for
achieving economic plantwide control. In the next Chapter, we consolidate the qualitative
discussions here into a systematic step-by-step procedure for synthesizing an economic
plantwide control system. The application of the procedure to five example processes with
rigorous dynamic simulation results is presented in the subsequent chapters.

Figure 12.5. Use of bottleneck constraint as TPM to reduce overrides in the two column
recycle process example

132
Figure 12.6. A control structure for the two column recycle process that allows operation at
V1MAX and V2MAX with no back-off

133
 MODULE IV

ECONOMIC PLANTWIDE CONTROL DESIGN

PROCEDURE AND CASE STUDIES

With an appreciation of the regulatory and economic considerations in plantwide control system
design, we are now ready to develop a systematic plantwide control system design procedure.
We develop and present such a design procedure, which is a natural extension of the pioneering
work of Page Buckley (DuPont), William Luyben (Lehigh), Jim Downs (Eastman) and Charlie
Moore (Tennessee). Its application to four realistic processes, namely, a recycle process with
side reaction, an ethyl benzene process, a cumene process and a C4 isomerization process is also
demonstrated. The last two examples are very comprehensive in that the performance of the
economic plantwide control structure synthesized from our procedure is compared with a
conventional plantwide control structure.

134
 Chapter 13. Systematic Economic Plantwide Control Design Procedure
With the preliminaries on regulatory and economic operation considerations in plantwide
control, we are now ready to develop a systematic procedure for designing an economic
plantwide control system for integrated chemical processes. For completeness, we review the
major contributors to plantwide control research before developing the procedure.
The design of effective plantwide control systems for safe, stable and economic process
operation of complex chemical processes with material and energy recycle has been actively
researched over the last two decades. The ready availability of dynamic process simulators has
been crucial in fostering the research. Over the years, Luyben and co-workers have done seminal
work in highlighting key regulatory control issues such as the snowball effect 15 in reactor-
separator recycle systems and suggesting practical control system structuring guidelines
(Luyben’s rules 16) for ensuring robust process stabilization in light of the same. Based on
several case-studies, a nine-step general procedure has been developed for synthesizing effective
plantwide control structures for integrated chemical processes 14. In their procedure, economic
concerns are addressed indirectly in the form of requiring ‘tight’ control of expected economic
variables such as product impurity, process yield etc. The control objectives are obtained using
engineering insights and heuristics.
Skogestad 24 has developed a more systematic steady state optimization based approach
for obtaining the control objectives. Typically, at the optimum steady state, multiple process
constraints are active so that these constraints must be controlled tightly. For managing the
remaining unconstrained steady state degrees of freedom, the control of self-optimizing
controlled variables 23 (CVs) is recommended. By definition, when self-optimizing variables are
held constant at appropriate values, near-optimal operation is achieved in spite of disturbances.
The quest for the best self-optimizing CV set is however not always straight-forward.
The combinatorial nature of the control structure design problem results in several
possible structures that provide safe and stable process operation. A very simple example is a
single-inlet single-outlet surge tank with two possible orientations for its level controller. In a
simple distillation column, assuming the feed is fixed, the two orientations each for the reflux
drum and bottom sump level controllers results in the well-known four basic regulatory control
configurations. Other control configurations are possible if instead of the process feed, one of the
other associated streams (distillate, bottoms, reflux or reboiler steam) is kept fixed. In a multi-
unit chemical process, there would clearly be several possible reasonable control configurations.
An obvious question then is which one is best for realizing economically (near) optimal process
operation with robust stabilization over the expected process operating space. Further, is there a
systematic methodology for synthesizing such an ‘optimal’ control structure?
A careful evaluation of the plantwide control literature reveals that most of the reported
case studies consider process operation around the design steady state (see these example case
studies 1,18,27), although more recently, also at maximum throughput 2,3,11,22. Around the base-
case design steady state, usually all the process units are sufficiently away from any capacity
constraints while at maximum throughput, typically, multiple units hit (hard) capacity
constraints. The active constraint set progressively expands with throughput to the full set at
maximum throughput. The expanding set partitions the throughput range into distinct regions.
Much of the open plantwide control literature addresses control system design only for a fixed
active constraint set, that is, only for a distinct region. This is surprising given that a plant must
be operated over a wide throughput range with different active constraints over its life-span.

135
 In this work, we develop a systematic approach for designing a simple and robust
plantwide control system for near-optimal process operation over a wide throughput range with
an expanding active constraint set. The approach has evolved out of very recent comprehensive
case-studies from our group 7-9. While the principles on which it is based may be well-known,
our main contribution is in bringing these scattered principles together into a meaningful, holistic
and practical top-down plantwide control system design framework. The application of the
proposed framework is demonstrated on three realistic example processes.

13.1. Degrees of Freedom (DOFs) and Plantwide Control Structures

The plantwide control system design problem may be viewed as seeking the best possible
way of managing the available control valves (control DOFs) for ensuring safe, stable and
economic process operation in the face of principal disturbances that include large changes in the
production rate (throughput) as well as variability in raw material quality, ambient conditions,
equipment characteristics and economic conditions (e.g. volatility in the energy prices etc). If we
discount the valves used to control nonreactive material inventories (surge tank levels, given
column pressures etc), the number of independent control valves remaining equals the steady
state operational DOFs for the process, which by definition, is the number of independent
specifications necessary to solve for the steady state solution. For a given process, one may use
alternative sets of independent specification variables. From the control perspective, each such
DOF specification variable is an independent CV (excluding non-reactive material inventory
controllers) in the plantwide control system. Note that one setpoint gets used to set the process
throughput and is referred to as the throughput manipulator (TPM).
Figure 13.1 provides an illustration of the one-to-one correspondence between the
independent CV setpoints (including TPM; excluding non-reactive material inventory
controllers) and the steady state DOF specification variable set for a simple reactor-recycle
process with five steady-state operation DOFs. The 5 DOFs are related to 1 fresh feed, 2 reactor
specifications (level and temperature) and 2 specifications for the column. Four alternative DOF
specification sets are shown in Figure 13.1. Implicit in each set is an inventory control system for
balancing of the process material and energy inventories as well as appropriate pairings for
controlling the specification variable. We have used the radiation rule 20 for material inventory
control which gives the orientation of the level controllers upstream and downstream of the TPM
respectively, opposite and in the direction of process flow, respectively. Note that for a given
DOF specification set, multiple possibilities exist for the choice of the pairings for controlling
the specification variables as well as for the inventory loops. Lastly, there exists flexibility in the
choice of the DOF specification variable set (CV set) itself. There thus exists tremendous
flexibility in designing the plantwide control system which must be gainfully exploited for
achieving the twin objectives of robust stabilization and economic operation.

136
 Figure 13.1. One-to-one correspondence between CV setpoints and steady state specification
variables for a simple recycle process
13.2. Two-Tier Plantwide Control System Design Framework

The control system of a process plant has two main objectives:

1. Optimal economic operation: Control economic CVs
2. Stable operation: Control drifting inventories (i.e. material balance control)
'Inventory' is interpreted here in its most general sense to include material, phase, component and
energy inventories in the different units as well as the overall process. The CVs for process
inventory regulation (material balance control) are usually obvious. They typically include liquid

137
levels and pressures, as well as selected temperatures, for example, a sensitive temperature in a
distillation column. The best CVs for economic operation at a given throughput may be obtained
from steady state optimization. Alternatively, process insight or operating experience may also
suggest economically sound CVs that should be controlled.
Optimal operation requires operating the process at the optimal point, that is, at all the
optimally active constraints as well as at the optimum value for decision variables corresponding
to any remaining unconstrained DOFs. Typically, multiple constraints are active at the optimum
solution. The choice of the unconstrained decision variable (CV) should be such that its optimum
value is relatively insensitive to disturbances, for example, in feed rate or composition. This is
the idea of 'self-optimizing' control where the economic loss due to no reoptimization for the
disturbance is acceptably small. Purely from the steady state operation perspective, a constant
setpoint operating policy with such CVs provides near-optimal operation in the face of
disturbances. In summary, the economic CVs for optimal operation are the active constraints at
the optimum plus the self-optimizing CVs corresponding to any unconstrained DOFs.
Once the set of economic CVs for a specified throughput are known (tier 1), either from
economic optimization or from heuristics, the economic and regulatory loop pairings must be
selected (tier 2). Which one of the two objectives (economic control or regulatory control)
should have priority when designing the control system pairings (structure)? In the commonly
used 'bottom-up' approach, process regulation is given priority over economic control. A 'basic'
or 'regulatory' control layer with focus on inventory control (stabilization), usually with the feed
rate as the throughput manipulator (TPM), is first designed. On top of this, one adds an
'advanced' or 'supervisory' control layer, often implemented using model predictive control,
which aims at achieving optimal economic operation by adjusting the setpoints into the
regulatory layer.
A problem with the 'bottom-up' approach is that it can yield slow control of the economic
variables due to unfavorable pairings, since control valves are already paired up for regulatory
control. This results in economic losses mainly because slow control requires back-off from hard
active constraint limits, which can be especially costly when it is optimal to maximize
throughput. As illustrated in Figure 13.2, the back-off and consequent economic penalty is
primarily determined by the severity of transients in the active constraint for the worst-case
disturbance. Even if the constraint is a soft one, tight regulation of the same may be desirable due
to the often very non-linear nature of the process with highly skewed deviations in only one
direction.

Figure 13.2. Illustration of tightness of active constraint control and back off

138
 In this work, we consider the alternative 'top-down' approach for selecting the control
pairings with higher priority to economic control over regulatory control. Such a reprioritization
is natural in light of the global push towards green / sustainable / efficient process operation. In
this approach, the best possible pairings for tight control of the economic CVs are obtained first
followed by pairings for inventory (material balance) control. It attempts to accomplish
economic and regulatory control in a single layer. The same is made possible as many-a-times
controlling an economic CV accomplishes a regulatory task (and vice versa). Also, processes are
designed to have sufficient number of surge capacities and the associated control valves remain
available for dynamic control (including inventory control) with no steady state economic
impact.
Regardless of the specific pairing philosophy (bottom-up or top-down), the application of
the two-tiered framework is relatively straightforward for a given active constraint set, implying
a fixed set of economic CVs that must be controlled. For most plants however, the active
constraint set expands or contracts depending primarily on the plant throughput. The best
economic CV set would then depend on the active constraint set (operating region) and conflicts
can arise with a control valve being most suitable for robust inventory control in one region and
economic CV control in another. Also, pairings done without considering the impact of a
constraint going active can result in loss of crucial control functions such as product quality
control or component inventory control with consequent snowballing. Additional override
controllers that alter the material balance control structure may need to be configured to ensure a
seamless transition and stable operation in the different regions. Alternatively, one can exploit
apriori knowledge of the full active constraint set to devise a plantwide control system that
ensures control of all critical economic and regulatory control objectives regardless of which
constraints in the full active constraint set are active. Such a control system is appealing in that
its basic regulatory structure remains fixed regardless of the operating region while also avoiding
the need for complex over-ride controllers. The two-tiered framework must be appropriately
modified to systematically devise such a control structure.

13.3. Active Constraint Regions for a Wide Throughput Range

A process is typically designed for a design throughput, where no hard constraints are
active due to over-design of the different processing units. Over its life span, economic
considerations necessitate sustained operation at throughputs much below and above the design
throughput, usually including operation at maximum achievable throughput. As throughput
increases above the design throughput, different processing units reach their (typically hard)
capacity constraints, usually one after the other. These active constraints partition the entire
throughput range into distinct regions. There are many disturbances in a plant, but throughput is
usually considered the principal disturbance because of its wide range encompassing multiple
active constraints. A control system that works well for such a large throughput range would also
handle other routine disturbances well.
Figure 13.3 illustrates active constraint regions with respect to throughput for a process
with 5 steady state DOFs. The active constraints divide the entire throughput range into three
regions corresponding to low (2 active constraints), intermediate (3 active constraints) and high
throughputs (4 active constraints). At the maximum achievable throughput (5 active constraints),
all the steady state DOFs are used up to drive as many constraints active in this hypothetical

139
example. Alternatively, one may have unconstrained DOFs remaining at maximum throughput
(i.e. throughput decreases on moving the unconstrained variable away from its optimum value).

Figure 13.3. Active constraint regions with respect to throughput

Let us assume that the full active constraint set, corresponding to maximum throughput
operation, does not change for a given process a. To design a truly top-down control system
where economic objectives are given the highest priority, loops for the tightest possible control
of all the active constraints would first be designed. We would then have the fewest number of
control valves left for process regulation, specifically material (total, component and phase) and
energy inventory control of the different units and the plant as a whole. If we can achieve
effective inventory regulation for maximum throughput operation along with the tightest possible
control of the economic CVs, the control system would most certainly work at lower throughputs
with additional DOFs (setpoints) available for control due to constraints becoming (optimally)
inactive. The reason we emphasize tight economic CV control at maximum throughput is that
this is where the economic benefits of improved operation are usually the largest.

13.4. Systematic Control System Design Procedure

Based on the above arguments, the two-tier plantwide control system design framework
is modified to designing a robust control system for process operation at maximum achievable
throughput with tight economic CV control, arguably the most difficult to stabilize due to the
highest number of active constraints, and then designing loops for taking up additional control
tasks using constraints (setpoints) that become optimally inactive at lower throughputs. The
additional control task may be economic CV control or throughput manipulation A step-by-step
'top-down' procedure for designing the overall control system for near optimum operation over a
wide throughput range is then:
Step 0: Obtain active constraint regions for the wide throughput range
Step 1: Pair loops for tight control of economic CVs at maximum throughput
Step 2: Design the inventory (regulatory) control system
Step 3: Design loops for ‘taking up’ additional economic CV control at lower throughputs
along with appropriate throughput manipulation strategy
Step 4: Modify structure for better robustness / operator acceptability
Each of these distinct steps is now elaborated upon.

a
This appears to be a reasonable assumption.

140
13.4.1. Step 0: Obtain active constraint regions for the wide throughput range

Steady state optimization of the available steady state DOFs is performed to obtain the
expanding set of active constraints with increasing throughput. A wide throughput range, from
below design throughput to the maximum achievable, is considered. The active constraints
partition the entire throughput range into distinct regions. To assess the economic impact of a
back-off in any hard active constraints, obtain the economic sensitivity of the hard active
constraints at maximum throughput, which corresponds to the full active constraint set. The
sensitivities dictate the prioritization as to which constraints must be controlled the tightest.
Corresponding to the unconstrained DOFs in an active constraint region (including
maximum throughput), propose self-optimizing CVs that give near-optimal operation with
constant setpoint. Sometimes such self-optimizing CVs are not forthcoming. This is acceptable
with the implicit understanding that these setpoints are adjusted by a real-time optimizer.

13.4.2. Step 1: Pair loops for tight maximum throughput economic CV control

The economic CVs at maximum throughput are all the active constraints (full active
constraint set) and self-optimizing CVs corresponding to any unconstrained steady state DOFs.
Typically constraints on maximum allowable product impurity, maximum allowable effluent
discharge etc. would be active along with hard capacity constraints such as column operation at
flooding limit, furnace operation at maximum duty etc. The full active constraint set may include
direct MVs (e.g. a fully open valve). Direct MVs that are optimally at a constraint limit should be
left alone at the limit and not used for conventional control tasks. Other active output constraints
should be selected as CVs and tightly controlled using close-by MVs that are not active
(saturated). For direct MV active constraints, the back-off is then eliminated while for active
output constraints, the back-off is mitigated by the tight control.
After implementing loops for tight active constraint control (including leaving a direct
MV at its limit), design loops for tight control of self-optimizing CVs. The economic optimum
with respect to these unconstrained variables is often 'flat' so that the economic penalty for small
deviations from the optimum setpoint is likely to be smaller than for a back-off from an active
constraint limit. The loops for self-optimizing CV control are therefore implemented only after
the loops for tight active constraint control. The flexibility in the input-output (IO) pairings then
gets utilized for the tightest control of the economically most important CVs.
There may be situations where the best self-optimizing CV exhibits extremely slow and
difficult dynamics. The control implementation may then be decomposed into a faster loop that
controls a dynamically better behaved close-by secondary CV, which is not the best self-
optimizing CV, with a cascade loop above adjusting its setpoint to ensure that the best self-
optimizing CV is maintained close to its (optimum) setpoint over the long-term.
We also note that economic optimality usually requires maximizing reactive inventory
hold up, for example, liquid (gas) phase reactor operation at maximum level (pressure). The best
pairings for tight control of these inventories should be implemented in this step itself with the
remainder of the inventory control system being synthesized in the next step (Step 2).

141
13.4.3. Step 2: Design the inventory (regulatory) control system

Given loops for tight economic CV control at maximum throughput, implement

appropriate loops for consistent inventory control 4 of the different units and the overall process.
Inventory is interpreted in its most general sense to include total amount of material, phases (e.g.
liquid or vapour), components as well as energy held within the individual units and the overall
process. Ensuring consistency of the inventory control system then accounts for tricky regulatory
plantwide issues such as the snowball effect due to the integrating nature of component
inventories in recycle systems. As recommended in Luyben et al. 14, 16, a ‘Downs Drill’ must be
performed to ensure the control system guarantees that no chemical component (and energy)
builds up within the process.
We note that processes are designed with sufficient number of surge capacities to
smoothen flow imbalances and facilitate start-up / shut-down. Thus, even if all steady state DOFs
are exhausted at maximum throughput to drive as many constraints active, these surge capacities
with their associated independent control valves ensure availability of control valves for
inventory regulation. An example is a simple distillation column with two steady state DOFs and
five control valves (excluding feed). Let us say that to minimize energy consumption, the light
key and heavy key in respectively the bottoms and distillate should be at their maximum limits.
The 2 steady state DOFs thus get exhausted in driving as many constraints active. If two valves
(e.g. reflux and reboiler steam) are paired for maintaining the light-key and heavy key impurities
in the two product streams at their maximum limits, three valves (e.g. distillate, bottoms and
condenser duty) remain available for controlling the three inventories (reflux drum level, bottom
sump level and column pressure).
In a top-down sense, inventory regulation (stabilization) is a lower objective than
economic control. The economic CV control loops are therefore put in place first (Step 1)
followed by the inventory control system (Step 2). In the inventory loops, local unit specific
pairings should be used \to the extent possible. However since valves already paired in Step 1 for
tight economic CV control are unavailable, some of the inventory loop pairings may possibly be
unconventional non-local 'long' loops.
It is important that, at least in the first pass, a truly 'top-down' plant-wide control structure
with such unconventional inventory loops be synthesized. In situations where the inventory
control turns out to be fragile due to these unconventional loops, the economic CV loop and
inventory loop pairings can always be appropriately revised (this is Step 4 of the procedure).
Many a times, these unconventional and seemingly unworkable inventory loops actually work
surprisingly well in practice. An example is bottom sump level control of a column with a very
small bottoms stream, akin to a leak compared to the internal column flows. Conventional
wisdom would suggest using such a leak stream for bottoms level control is unworkable and
therefore ill-advised. If however a stripping section tray temperature is well controlled e.g. by
adjusting the boilup or feed, the seemingly unworkable pairing provides acceptable sump level
control 25. Level control would be lost only when the temperature loop is put on manual. In our
opinion, the unconventional level controller pairing is acceptable with the caveat that the
stripping temperature loop be viewed as part of the overall inventory control system and never
put on manual. One of the case-studies provides another example where an unconventional
inventory control loop pairing works surprisingly well.

142
13.4.4. Step 3: Design loops for additional economic CV control at lower throughputs along
with throughput manipulation strategy

In the control structure for process operation at maximum throughput, one setpoint
(TPM) must be used to reduce the process throughput below maximum. Usually, the setpoint for
the last constraint to go active is an immediate choice for the TPM. Moving this TPM setpoint
away from its active constraint limit would reduce the throughput. As throughput is reduced,
additional active constraints become optimally inactive, typically, one after the other. The
unconstrained setpoints of the corresponding constraint controllers are now MVs that may be
used to control additional self-optimizing CVs for near-optimal operation at lower throughputs.
For dynamic reasons, the new CV should be close to the MV (constraint controller setpoint) that
becomes available. If such a close-by pairing is not forthcoming, the new unconstrained setpoint
may alternatively be considered for use as the TPM in that active constraint region, while using
the 'old' TPM (from the more constrained higher throughput region) to control the new CV. The
best throughput manipulation strategy across the wide throughput range would then depend on
the specific full active constraint set.
To develop such a scheme, list the MV setpoints that become unconstrained along with
close-by CVs whose control can be taken-up for more economical operation. Usually,
conventional control tasks are best taken up by these MV setpoints. An example is a column
moving away from its flooding limit and the resulting unconstrained boilup (MV) taking up
column tray temperature control for better energy efficiency. In this list, the unconstrained MV
setpoint that gives the dynamically poorest economic CV control may be used as the TPM. In the
special case where this MV setpoint is the last constraint to go active and its optimal variation
with throughput is monotonic, this single setpoint can be used as the TPM over the entire
throughput range. If optimality requires holding this MV setpoint constant in a lower throughput
region, the TPM must be shifted to the setpoint of the constraint variable that becomes inactive
in that lower throughput region. The shifting may have to be repeated depending on the nature of
the next constraint that goes inactive on decreasing throughput.
Referring back to Figure 13.3, we note that the next constraint to become active as
throughput is increased can always be used as the TPM in that operating region. If we keep
shifting the TPM to the next constraint to go active as throughput is increased, the back-off from
the active constraint limit is mitigated. In particular, using the unconstrained setpoint of a
constraint control loop as the TPM allows the setpoint to be left closest to its active limit with the
least back-off. If the constraint is economically dominant (i.e. large economic penalty per unit
back-off), both throughput manipulation and reduced economic penalty due to mitigated back-off
get achieved. Another pairing possibility that allows the same is using the unconstrained setpoint
of the constraint control loop to control a self-optimizing CV, and not a critical CV such as
product quality (critical for economic reasons) or a process inventory (critical for process
stabilization). When the constraint limit is reached (e.g. when throughput is increased), control of
the non-critical self-optimizing CV is simply given up and the constraint variable setpoint is left
closest to the constraint limit with the least back-off. In the special case where the active
constraint is a saturated valve, the valve gets left at its saturated position with no back-off.
The point is that there is nothing sacrosanct about fixing the TPM location, although it
may be desirable that operators have a single handle to adjust the throughput. This flexibility
should be gainfully exploited for eliminating / mitigating the back-off in economically dominant
active constraints, obtaining pairings for tight control of the additional unconstrained economic

143
CVs at lower throughputs as well as simplifying the overall plantwide control system. The
throughput manipulation strategy is therefore best considered along with the additional
unconstrained economic CV loop pairings in a single step. The best throughput manipulation
strategy usually becomes self evident in light of the particular full active constraint set.

13.4.5. Step 4: Modify structure for better robustness / operator acceptance

The control structure obtained from Step 1-3 corresponds to a fully top-down design
approach where tight economic CV control at maximum throughput is given precedence over
regulatory inventory control, for which control valves are typically available by the design of the
process. Through carefully chosen input-output (IO) pairings, the structure attempts to transform
all the process variability to the surge capacities and utilities, while maintaining economic CVs
at their constrained / optimum setpoints. In such a structure, we may have inventory control
loops that are quite unconventional with long loops across units. These may result in fragile
inventory (including energy inventory) control.
A surge drum overflowing or drying for even moderately large flow disturbances is a
typical result of inventory control fragility. Another example is temperature control of a highly
exothermic CSTR with maximum reactor cooling duty being an active constraint. If the cooling
duty is left alone at maximum (as it is active) and the CSTR temperature is controlled using the
reactor feed, there is the possibility of a thermal runaway with reactants slowly building up
inside the reactor when the temperature is below setpoint and the accumulated reactants lighting
up once the temperature starts to rise back-up due to the exponential dependence of reaction rate
on temperature. The energy inventory inside the reactor then blows up, which is unacceptable.
The IO pairings must then be revised to improve inventory control robustness.
To revise the pairings, in the control structure obtained for maximum throughput
operation (Step 1-3), tight control of one or more economic CVs must first be given up to free
appropriate control valves that then get paired for robust / conventional inventory control. The
valves (or setpoints) that become available in lieu may be used for less tight or loose control of
the economic CVs whose control was earlier given up. In this exchange of economic CV and
unconventional inventory loop MVs for a more robust / conventional inventory control system, it
is preferable that the economic CV with the least economic impact (lowest sensitivity) be used to
minimize the economic penalty. Instead of unconventional 'long' inventory loops, the revised
structure would then have more conventional inventory loops with 'long' economic CV loops.
In most chemical processes, only a few active constraints are dominant with a large
economic penalty per unit back-off. With appropriate iteration between Step 1-3, it should be
possible to synthesize a control system for tight control of the few dominant active constraints
with a not-too-unconventional (i.e. acceptable) and robust inventory control system along with
well-behaved additional unconstrained economic CV loops at lower throughputs.
The application of the systematic approach for economic plantwide control system design
is demonstrated on four realistic process examples. The first example process is a hypothetical
reactor-separator-recycle process with side reaction. The second example process is a C4
isomerization process. The ethyl benzene manufacturing process is the third example considered.
We finally consider two alternative processes for cumene manufacture.

144
 Chapter 14. Economic Plantwide Control of Recycle Process with Side
Reaction
14.1. Process Description

The process flowsheet is shown in Figure 14.1 and consists of a cooled liquid phase
CSTR followed by a stripper and a distillation column. The main reaction A + B
C and the
minor side reaction C + B
D occur in the CSTR. Reaction kinetics and other modelling details
are available in Jagtap et al. 7. The unreacted A and B in the reactor effluent are stripped,
condensed and recycled along with some C. The stripper bottoms is fractionated to recover 99%
pure C as the distillate (main product) and D with some C as the bottoms (side product). The
process has 7 steady state DOFs (2 fresh feeds, reactor level and temperature, 1 stripper DOF and
2 column DOFs) and there are 13 independent control valves. Thus even if all steady state DOFs
are exhausted at maximum throughput, 6 valves would still remain available for dynamic
control, including inventory control.

Figure 14.1. Schematic of recycle process with design and base operating conditions
145
 14.2. Economic Plantwide Control System Design

Table 14.1 neatly summarizes the step-by-step implementation of the four-step economic
plantwide control system design procedure to this process. A reasonably detailed explanation of
the steps is provided in the following.

Table 14.1. Economic Plantwide Control Structure Synthesis for Recycle Process
Step 0: Active Constraint Regions and Economic CV’s
Region I II III Max Throughput
MAX MAX
Additional Active V1 V1 TRxrMAX
- V1MAX TRxrMAX V2MAX
Constraints*
Unconstrained DOF’s 2 1 0 0
Self-Optimizing CV’s xBRxr, TRxr xBRxr - -

Step 1: Maximum Throughput Economic Control Loops

Active Constraint TRxrMAX↔QRxr V1MAX↔QReb1 V2MAX↔QReb2 TSCol↔B1

Control Loops xBColD↔TStpSP ↔ FStpSP xDColD↔L2/B1SP↔L2SP LVLRxrMAX↔FTotRxr↔FA
Self-Optimizing Loops none
Step 2: Maximum Throughput Inventory Loops
LVLReb2↔ B1 LVLCnd1↔FRcy PCnd1↔QCnd1
LVLReb1↔ xBRxr SP↔(FB/FTotRxr)SP↔FB LVLCnd2↔D2 PCnd2↔QCnd2
Step 3: Additional Self-Optimizing CV Loops at Reduced Throughput
Region III Region II Region I
TPM: V2SP
TPM: V2SP
TRxrSP #↔QRxr
TPM: V2SP xBRxr SP #↔(FB/FTotRxr)SP↔FB
xBRxr SP #↔(FB/FTotRxr)SP↔FB
LVLReb1↔TRxrSP↔QRxr
LVLReb1↔V1SP↔QReb1
Step 4: Modifications for Conventional Inventory Control Loop
LVLReb1 ↔ B1; TSCol2 ↔ V2SP(with sufficient back-off in V2SP)
Region III Region II Region I
TPM: V1SP
TPM: TRxrSP
TPM: xBRxr SP TRxrSP #
xBRxr SP #
xBRxr SP #
*: LVLRxrMAX, xBColD, xDColD, TSCol are always active; #: Set point value is the optimized value

146
14.2.1. Step 0: Active Constraint Regions and Economic Operation

To avoid product give-away, the product C impurity mol fractions are fixed at their
specified upper limits of 0.98% B (xBColD) and 0.02% D (xDColD) for the desired 99 mol% pure C
(xCColD) product. At maximum throughput, the active constraints are maximum column boilup
(V2MAX), reactor temperature (TRxrMAX), stripper boilup (V1MAX) and reactor level (LVLRxrMAX).
Further, to prevent loss of precious C with the side product, the average temperature of three
adjacent sensitive stripping trays (TSCol) is maintained a. The four equipment capacity constraints,
the two product impurity mol fractions and the product column stripping section temperature
specification exhaust all 7 steady state DOFs.
At lower throughputs, it is economically near optimal to hold the two product impurity
mol fractions and the column stripping section temperature at their maximum throughput values.
Also, the LVLRxrMAX constraint is active at all throughputs as it maximizes the reaction
conversion at a given reactor temperature. As throughput is reduced below maximum, the
capacity constraints become optimally inactive in the order V2MAX, TRxrMAX and V1MAX. The
entire throughput range thus gets partitioned into three active constraint regions (see Table 14.1,
Step 0). The number of unconstrained steady state DOFs corresponding to the low throughput
(only LVLRxrMAX active), intermediate throughput (LVLRxrMAX and V1MAX active) and high
throughput (LVLRxrMAX, V1MAX and TRxrMAX active) regions is respectively, 2, 1 and 0. The
V2MAX constraint going active represents the loss of DOF corresponding to specifying the
throughput. The process throughput is then determined by the actual 7 equality / inequality
constraint variable values. Jagtap et al. 11 have shown that in the low throughput region, holding
the reactor temperature (TRxr) and the CSTR inlet B (limiting reactant) concentration (xBRxr) at
appropriate constant values provides near-optimal steady operation. In other words, TRxr and
xBRxr are self-optimziing CVs corresponding to the two unconstrained DOFs. In the intermediate
throughput region, holding xBRxr constant ensures near optimum steady operation (TRxrSP is not
held constant and adjusted for either active constraint control or throughput manipulation). In the
high throughput region, there are no unconstrained steady state DOFs left.

14.2.2. Step 1: Loops for Tight Control of Full Active Constraint Set

We now design the control system for maximum throughput operation, where all
constraints in the full active constraint set are active. At maximum throughput, there is no TPM
as all steady state DOFs are exhausted implying the DOF related to throughput is used for active
constraint control. V2MAX and V1MAX are active hard constraints with significant economic
penalty. Any back-off from V2MAX causes a large loss in throughput and any back-off in V1
causes a reduction in the recycle rate and hence a loss in selectivity. Accordingly, V1 and V2 are
controlled tightly using the respective reboiler steam valves. The back-off necessary from V1MAX
and V2MAX is then almost negligible.
It is economically important to have tight control of the impurities in the product. The
product impurity D mol fraction (xDColD) is controlled using the column reflux. The composition
controller manipulates the reflux-to-feed ratio setpoint b. Maintaining product impurity B mol
a
This ensures that C composition in the byproduct stream remains small
b
In practice, the composition controller would cascade a setpoint to a rectifying tray temperature controller which
manipulates the L/F ratio setpoint.

147
fraction (xBColD) requires tight control of the B dropping down the stripper as all of it ends up in
the product. Since V1MAX is active, V1 cannot be used for stripper tray temperature control. The
stripper temperature (TStp) controller then manipulates the stripper feed (FStp), which provides
tight temperature control. The temperature setpoint is adjusted by a cascade xBColD controller.

LVLRxrMAX and TRxrMAX, the other active equipment capacity constraints imply LVLRxr
and TRxr must be controlled tightly. Controlling LVLRxr and TRxr (at their maximum limits)
would also stabilize the reactor material and energy inventories, respectively. For tight control,
TRxr is controlled using reactor cooling duty (QRxr), the MV with the best dynamic response (fast
dynamics and high open loop gain). We assume TRxrMAX to be a soft constraint and set TRxrSP =
TRxrMAX. The orientation of the reactor level controller must be opposite to process flow since the
reactor effluent (FStp) is already paired for stripper temperature control. The total flow to the
reactor (FTotRxr) is a good MV for tight reactor level control. Accordingly, LVLRxr is controlled
by adjusting FTotRxr SP, which in turn is maintained by manipulating the fresh A feed (FA).
Lastly, it is economically important to maintain an appropriate column stripping section
temperature (TSCol) to ensure loss of precious C in the bottoms is kept small. The active V2MAX
constraint implies column boilup is unavailable for temperature control. Accordingly, the column
feed (B1) is manipulated for the purpose. The active constraint control loops are shown in Figure
14.2. The constrained setpoints at maximum throughput are highlighted in brown.

Figure 14.2. Plantwide control structure for maximum throughput operation of recycle
process (Case Study I)

148
14.2.3. Step 2: Inventory (Regulatory) Control System

Control loops to stabilize the liquid, vapour and component inventories in the process are
now implemented using the available unpaired valves (reactor level and energy is already
stabilized by the LVLRxr and TRxr loops). The inventory loops are shown in blue in Figure 14.2.
We need to control the column reflux drum and sump levels, the stripper sump level and the
recycle condenser level. The column and the recycle condenser pressures also need to be
controlled.
The existing loops for tight active constraint control in Figure 14.2 imply obvious loop
pairings for inventory control. The column reflux drum level (LVLCnd2) is controlled using the
distillate (D1). The recycle and column condenser pressures (PCnd1 and PCnd2) are controlled using
the respective cooling duty valves (QCnd1 and QCnd2). The column sump level (LVLBot) is
controlled using the feed from the stripper (B1). To mitigate transients in the reactor composition,
FB is maintained in ratio with FTotRxr. To ensure A or B component inventory does not build up
inside the recycle loop (snowball effect), the B mol fraction in the reactor inlet (xBRxr) is
maintained by adjusting the FB to FTotRxr ratio setpoint (FB/FTotRxr SP).
With these pairings, no close-by valves are left for controlling stripper sump level
(LVLStp). The only available option is to adjust the xBRxr SP. The pairing makes sense in that the
reaction products accumulate in the stripper sump for downstream separation. The sump level is
then an indirect indication of the reactor production rate. If this level is falling, the reactor
production needs to be increased. Increasing the xBRxr SP causes the limiting reactant B
composition in the reactor to increase with consequent increase in generation of product C and
hence in the stripper sump level.
The stripper level controller is the most unconventional in the scheme. Will it work in
practice? That depends on the hold up in the CSTR. If the reactor is too big, the dynamic effect
of a change in the xBRxr SP on stripper sump level would be slow and it may run dry or overflow
during worst case transients. The robustness of the control system is tested for a ±5 5% step bias
in the FB sensor (control system tuning details in Appendix A). In the transient response, all the
levels are well controlled with the maximum deviation in the stripper sump level being < 4%.
The inventory control scheme, though unconventional, is quite robust and acceptable.

14.2.4. Step 3: Additional Economic CV Control Loops and Throughput Manipulation

At lower throughputs, the additional unconstrained economic CVs whose control must be
taken up are xBRxr and TRxr. Both are associated with the reactor. Since maximum column boilup
(V2MAX) is the last constraint to go active and its optimal variation with throughput is monotonic,
we consider using it as the TPM over the entire throughput range. Now as V2SP is reduced below
V2MAX, the production rate would decrease below maximum with xBRxr reducing. The excess A
inside the reactor then increases to further suppress the side reaction for improved yield to the
desired product. When xBRxr reduces to its optimal value, it must be held constant for optimal
operation. LVLStp then gets controlled using TRxrSP, in lieu of xBRxr. TRxrSP would reduce below
TRxrMAX as V2SP is decreased. When TRxrSP decreases to its optimum value, it must be held
constant. LVLStp then gets controlled using V1SP in lieu of TRxrSP). V1SP would reduce below
V1MAX as V2SP is reduced to decrease the throughput. The stripper bottom sump level controller
pairing thus switches from xBRxr SP to TRxrSP to V1SP as throughput is reduced. Referring to the
throughput regions in Table 14.1, at high throughputs, xBRxr floats to the appropriate value

149
determined by V2SP via the action of the inventory control system. At intermediate throughputs,
xBRxr is maintained at its optimum and TRxr floats to the appropriate value. Finally, at low
throughputs, xBRxr and TRxr are held at their near optimum values and V1SP floats to the
appropriate value.
A simple override scheme to accomplish the switching between the operating regions
with three separate PI stripper sump level controllers (LC1, LC2 and LC3) is shown in Figure
14.2. The MVs for LC1, LC2 and LC3 are respectively, V1SP, TRxrSP and xBRxr SP. At maximum
throughput, since TRxrMAX and V1MAX are active, LC1 and LC2 are inactive and sump level control
is performed by LC3. As V2SP (TPM) is reduced below V2MAX, LC3 decreases xBRxr SP. When
xBRxr SP reduces below its optimum value, the high select block, HS3, passes the optimum value to
the xBRxr controller. LC3 then becomes inactive and stripper sump level control is lost. The level
then increases beyond LC2 setpoint and the LC2 output starts to decrease. When the output
decreases below TRxrMAX, level control is taken over by LC2. When TRxrSP decreases below its
optimum value, the high select block, HS2, passes the optimum value and LC2 becomes inactive
and the stripper sump level again rises beyond LC1 setpoint. LC1 output then reduces and on
decreasing below V1MAX, the low select block, LS1, causes LC1 to take over level control. A
complementary logic causes proper switching from LC1 to LC2 to LC3 as throughput is
increased.
Note that the decreasing level setpoint order (LC1 > LC2 > LC3) is necessary to enforce
the proper switching order. For example, when LC1 is active, the level would be close to LC1
setpoint and the I action in LC2 and LC3 would cause the respective controller output signals to
be sufficiently high ensuring the respective (high) select blocks pass the appropriate signal
(optimum TRxrSP and xBRxr SP respectively). It is also highlighted that in the given scheme, LC1 is
reverse acting and nested with the stripper temperature loop. As LVLStp decreases, V1SP increases
(reverse action) which causes the stripper temperature to increase. The temperature controller
then increases the stripper feed which causes the LVLStp to return to setpoint.
Rigorous dynamic simulations are performed to test the synthesized control structure in
Hysys. Unless specified otherwise, all flow / pressure PI controllers are tuned tight for a fast and
snappy servo response. The non-reactive level controllers are P-only with a gain of 2. The only
exception is the unconventional stripper sump level controller with overrides. For the three
different pairings in the three operating regions, distinct conservative (non-aggressive) tunings
are used to dampen flow variability. The CSTR level is controlled using a PI controller for offset
free level tracking. The approximate controller tuning is first obtained using the Hysys autotuner
and then adjusted for a fast and not-too-oscillatory servo response at maximum throughput. All
temperature measurements are lagged by 2 mins to account for sensor and cooling / heating
circuit dynamics. To tune the temperature loops, the open loop step response at maximum
throughput is obtained and the reset time set to 1/3rd of the approximate 95% response
completion time. The gain is then adjusted for a slightly underdamped servo response with mild
oscillations. The composition controllers are similarly tuned. A sampling time and delay time of
5 mins each is applied to all composition measurements. Salient controller parameters are
reported in Table 14.2.
The dynamic response of salient process variables of this control system to a throughput
transition from the base-case throughput (FA = 100 kmol/h) to the maximum throughput (FA =
188.7 kmol/h) and back is shown in Figure 14.3. Tight product purity control is achieved along
with smooth plantwide transients. The control system is also tested for a ±5% step bias in the FB
measurement signal at maximum throughput operation. The dynamic response is plotted in

150
Figure 14.4. Notice the tight control of the product impurities as well the C loss in the by-product
stream. The synthesized plantwide control system is thus suitable for economic process operation
across the wide throughput range.
If a conventional control system with the TPM at the fresh feed were to be implemented,
the need for a back-off from V1MAX and V2MAX during worst case transients results in significant
throughput (economic) loss (~4-7%) 8. The synthesized plantwide control system thus achieves
significantly superior economic operation for the same plant equipment.

Table 14.2. Salient controller tuning parameter for recycle process

CV KC τi (min) Sensor Span
Rxr
xB 0.8 400 0–1
TRxr* 1 10 60 – 130 ºC
LVLRxr 0.5 25 0-100%
TStp 0.5 15 100 – 160 ºC
TSCol 0.6 25 140 – 180 ºC
xBColD 0.1 40 0 – 0.02
xDColD 0.1 30 0 0.0004
Tuning for LVLReb1 override control
LVLReb11 0.8 200 0-100%
LVLReb12 0.6 250 0-100%
LVLReb13 0.5 400 0-100%
All level loops use KC = 2 unless otherwise specified
Pressure/flow controllers tuned for tight control
All composition measurements: deadtime = 5 min; sampling time = 2 min;
*: Derivative action used with τD = 2 min
All temperatures measurements lagged by 2 mins
1: MV= V1; 2: MV= TRxr,; 3: MV= xBRxr

14.2.5. Step 4: Modifications for a More Conventional Inventory Control System

Given that the control system works well with the unconventional stripper bottoms level
control loop, Step 4 (control system modification for a more conventional inventory control
system) is not necessary. It is however instructive to develop a control system with conventional
local inventory control loops.
The stripper sump level control loop in Figure 14.2 is arguably the most controversial
inventory control loop. For a more conventional local pairing, the column stripping section
temperature (TSCol) loop is broken to free the stripper bottoms valve, which is then paired to
control the stripper sump level. TSCol may then be maintained by adjusting xBRxr SP in a long loop.
Even as the steady state economic penalty with such a long economic loop is small, the penalty
during transients is likely to be severe. Due to the V2MAX active constraint, the precious C that
could not be boiled off would accumulate at the bottom of the product column and get
discharged in the by-product stream by the action of the column sump level controller. Since the
optimum C leakage in the bottom stream is very small to begin with, one would expect transient
deviations in the direction of higher than optimum C leakage to be significantly more severe than
in the opposite (lower than optimum C leakage) direction, where there is little / no leeway. The
long column stripping section temperature loop is then susceptible to large loss of precious C
during transients. To mitigate the same, a local temperature control loop is needed. Accordingly,

151
TSCol is controlled using the column boilup (V2SP). For maximum throughput operation without
loss of control of C leaking down the product column bottoms, the xBRxr SP would be set at a
value such that V2MAX constraint is just hit during the worst case transient. The back-off from
V2MAX then represents an unrecoverable economic loss, which is the price that must be paid for a
more conventional inventory control system.
In the original control system (Figure 14.2), V2SP was used as the TPM in all regions.
With the revised pairings where V2SP is used for TSCol control, an alternative throughput
manipulation strategy is needed. To reduce throughput below maximum (Region III), xBRxr SP
gets used as the TPM. Once xBRxr SP is reduced to its optimum value, the TPM shifts to TRxrSP
which is reduced below TRxrMAX (Region II). Once TRxrSP is reduced to its optimum value, the
TPM shifts to V1SP, which is reduced below V1MAX (Region I). Note that in this TPM shifting
scheme, the back-off from V1MAX is negligible. Also, the transient variability in TRxr for
operation at TRxrMAX is minimal as TRxrSP is not adjusted by any master cascade loop once
TRxrMAX is hit. The revised control system is shown in Figure 14.5 (Step 4 in Table 14.1).

MAX achievable throughput

TRxr MAX

V1MAX V2MAX

Figure 14.3. Throughput transition with stripper sump level override control scheme

152
 TRxrMAX

V1MAX V2MAX

Figure 14.4. Transient response for ±5% step bias in FB flow sensor
—: +5% bias; —: -5% bias

153
Figure 14.5. Recycle process modified control structure for conventional inventory control
system

154
 Chapter 15. Economic Plantwide Control of Ethyl Benzene Process
15.1. Process Description

The process consists of two reactors and two columns along with two liquid recycle
streams, as in Figure 15.1. The reaction chemistry consists of three reactions
C6H6 + C2H4
C8H10 Main Reaction
Benzene Ethylene Ethyl
Benzene

C8H10 + C2H4
C10H14 Side Reaction

Ethyl Ethylene Diethyl
Benzene Benzene

C10H14 + C6H6
2 C8H10 Transalkylation

Diethyl Benzene Ethyl
Benzene Benzene

The reaction kinetics and other modeling details are available in Jagtap and Kaistha 8. The first
two reactions occur primarily in the first coil cooled CSTR while transalkylation primarily
occurs in the second adiabatic CSTR. Near complete ethylene conversion occurs in the two
CSTRs. The reaction section effluent is fractionated in the recycle column to recover and recycle
unreacted benzene back to the first CSTR. The bottoms is fractionated in the product column to
recover 99.9 mol% pure ethyl benzene (EB) as the distillate. The diethyl benzene (DEB) drops
down the bottoms and is recycled to the second CSTR. The DEB is allowed to build in the
recycle loop so that the DEB formation rate by the side reaction exactly matches the DEB
transalkylation rate for no net DEB formation. The DEB is thus recycled to extinction.

15.2. Economic Plantwide Control System Design

The step-by-step synthesis of the economic plantwide control system is summarized in

Table 15.1. The major steps are briefly described below.

15.2.1. Step 0: Active Constraint Regions and Optimal Operation

With fixed pressures, the process has nine steady state degrees of freedom: 2 fresh feeds,
2 DOFs for the first reactor (level and temperature), 1 for the second reactor (level) and 4 DOFs
for the two columns. At maximum throughput, there are 8 active constraints: maximum recycle
column boilup (V1MAX) and reflux (L1MAX), maximum product column boilup (V2MAX), first
reactor maximum temperature (Trxr1MAX) and level (LVLrxr1MAX), second reactor maximum level
(LVLrxr2MAX) plus maximum product impurity levels xBzD2 MAX (benzene mol fraction) and
xDEBD2 MAX (DEB mol fraction) for no product give-away. This leaves one unconstrained steady
state DOF at maximum throughput, which is related to the optimal DEB recycle (L1MAX fixes
benzene recycle). Of the active constraints, Trxr1MAX, LVLrxr1MAX and LVLrxr2MAX are active
regardless of throughputs. As throughput is increased, L1MAX, V2MAX and V1MAX become active,

155
in that order. These three active constraints are treated as hard while the remaining ones are
treated as soft.
In this process, unlike previous examples, an unconstrained DOF remains at maximum
throughput. The DEB recycle flow rate (B2) is considered as a self-optimizing CV. We have
shown that holding B2 fixed at its optimal maximum throughput value results in only a maximum
0.35% operating profit loss at lower throughputs 8. The loss is deemed acceptable and is a
consequence of energy being significantly cheaper than products or raw material (Douglas'
doctrine 5). At lower throughputs, overrefluxing in the two columns is mitigated by maintaining
L1 in ratio with the recycle column feed (Fcol1) and maintaining a sensitive stripping tray
temperature (TScol2) using V2. The self-optimizing CVs corresponding to unconstrained L1 and
and V2 are L1/Fcol1 and TScol2 respectively.

Figure 15.1. Schematic of ethyl benzene process with design and operating conditions

15.2.2. Step 1: Loops for Maximum Throughput Economic CV Control

The full active constraint set consists of LVLrxr1MAX, Trxr1MAX, LVLrxr2MAX, L1MAX, V2MAX,
MAX
V1 xDEBD2 MAX and xBzD2 MAX. Of these, L1MAX, V2MAX and V1MAX are hard constraints. For
negligible back-off from their hard constraint limits, V1 and V2 are controlled using the
respective reboiler steam valves (Qreb1 and Qreb2) while L1 is flow controlled. Trxr1MAX is
controlled using the reactor cooling duty (Qrxr), a conventional pairing for tight temperature
control. For tight control of xDEBD2 (product impurity), the column reflux to feed ratio is adjusted.
For tight control of xBzD2 (product impurity) another cascade loop arrangement is implemented
where the composition controller adjusts a sensitive recycle column stripping tray temperature
controller setpoint, which in turn manipulates the column feed (Fcol1). With the recycle column

156
feed (FCol1) paired for temperature control, the level controllers in the two reactors must be
oriented opposite to the process flow. Accordingly, LVLrxr2 is controlled using its feed (Frxr2).
Similarly, for tight level control of the first reactor (LVLrxr1), the reactor liquid feed (fresh +
recycle benzene, FTotBz) is adjusted. FTotBz is maintained by adjusting the fresh benzene so that
that the fresh benzene is fed as a make-up stream (Luybens' rule). Lastly, B2 (self optimizing
CV) is flow controlled.

Table 15.1. Economic Plantwide Control Structure Synthesis for Ethyl Benzene Process
Step 0: Active Constraint Regions and Economic CV’s
Region I II III Max Throughput

Additional Active V1MAX V2MAX

Constraints* - L1MAX V2MAX L1MAX
L1MAX

Unconstrained DOF’s 3 2 1 1
col2 col2
Self-Optimizing CV’s B2, L1/F , 1 TS B, 2 TS B2 B2
Step 1: Maximum Throughput Economic Control Loops

Active Constraint Trxr1MAX ↔Qrxr1 V1MAX↔Qreb1 V2MAX↔Qreb2 LVLrxr1MAX↔FTotBz ↔FBz

Control Loops xBzD2↔ TScol1 SP↔Fcol1SP xDEBD2↔L2/B1SP↔L2SP LVLrxr2MAX↔Frxr2

Self-Optimizing Loops none

Step 2: Maximum Throughput Inventory Loops
LVLcnd1↔D1 LVLreb1↔ FC2/FTotBz SP↔ FC2 Pcnd1↔Qcnd1

LVLcnd2↔D2 LVLreb2↔B1 Pcnd2↔Qcnd2

Step 3: Additional Self-Optimizing CV Loops at Reduced Throughput

Region III Region II Region I
TPM: V1SP
SP
TPM: V1
TPM: V1SP TScol2↔V2SP #
TScol2↔V2SP #
L1/F1↔L1#

Step 4: Modifications for Conventional LVLReb1 Control Loop

LVLreb1 ↔ B1
Region III Region II Region I
TPM: V1SP
TPM: V1SP
TPM: V1SP TScol2↔V2SP#
TScol2↔V2SP #
B2↔FTotBz/FC2SP B2↔FTotBz/FC2SP↔FC2
B2↔FTotBz/FC2SP #
L1/F1↔L1#
*: Trxr1MAX, LVLrxr1MAX, LVLrxr2MAX, xBzD2 MAX, xDEBD2 MAX are always active; #: Is unconstrained from MAX limit

157
15.2.3. Step 2: Inventory (Regulatory) Control System

The remaining inventories to be controlled include the four column levels (LVLcnd1,
LVLcnd2, LVLbot1, LVLbot2) and the two column pressures (Pcnd1 and Pcnd2). The column pressures
are controlled conventionally using the respective condenser duty valves (Qcnd1 and Qcnd2). The
reflux drum levels of the two columns (LVLcnd1 and LVLcnd2) are controlled using the
respective distillate stream (D1 and D2). On the product column, since the B2 is under flow
control as a self-optimizing variable and therefore unavailable, the sump level (LVLbot2) is
controlled using the product column feed (B1). This leaves no close-by valves for controlling
the recycle column sump level (LVLbot2). The only pairing possibility is to adjust the fresh
ethylene feed rate (FC2). To mitigate the transients in the reactor composition, FC2 is maintained
in ratio with the FTotBz with the LVLbot2 controller adjusting the ratio setpoint, FC2/FTotBz SP. As in
the recycle process case study (Case Study 1), this is an unconventional long inventory loop and
makes sense in that the reaction products (EB and DEB) accumulate in the bottom sump of the
recycle column. LVLBot2 thus indirectly indicates the production rate. A decreasing level implies
the reaction production rate must be increased, which is accomplished by increasing FC2 (limiting
reactant) via appropriate adjustment in FC2/FTotBz SP by the level controller.

15.2.4. Step 3: Additional Economic CV Loops and Throughput Manipulation

To reduce throughput below maximum, we consider using V1SP as the TPM across the
entire throughput range as V1MAX is the last constraint to go active. When optimally inactive,
L1SP is maintained in ratio with the recycle column feed to mitigate overrefluxing in the recycle
column e. Similarly, V2SP takes up tight control of a sensitive product column stripping tray
temperature, whenever feasible at lower throughputs.

15.2.5. Step 4: Modifications for a More Conventional Inventory Control System

The economic plantwide control structure synthesized by the application of Step 1-3 of
our procedure is shown in Figure 15.2. In this control system, we have an unconventional and
long loop for controlling the recycle column sump level. For this process, the total reactor
residence time is ~2 hrs so that the dynamic response of LVLbot2 to a change in FC2/FTotBz SP
(MV) is quite sluggish resulting in the recycle column sump overflowing or running dry even for
the mildest of disturbances such as a 1% step change in B2SP. Clearly the inventory control
system is very fragile so that the economic CV and inventory loop pairings must be appropriately
revised.
To revise the pairings, we first consider giving up on tight control of the self-optimizing
CV, B2. The product column sump level (LVLbot2) is then paired with B2 which frees up the
recycle column bottoms flow (B1) which is then used for robust control of LVLbot1. This frees up
FC2/FTotBz SP which takes up 'loose' control of the self-optimizing variable, B2. The long inventory
loop, LVLbot1 - FC2/FTotBz SP, in Figure 15.2 (Step 2 row in Table 15.1) thus gets replaced by a
long B2 - FC2/FTotBz SP loop after the re-pairing exercise to provide a conventional and robust
inventory control system. The revised control system is shown in Figure 15.3.

e
Alternatively, L1SP can take up rectifying temperature control for dual ended control.

158
 To transition to lower throughputs, V1SP, the last constraint to go active is used as the
TPM over the entire throughput range. Also, to prevent overrefluxing in the two columns at low
throughputs, V2SP takes up product column stripping tray temperature control and L1 is
maintained in ratio with the recycle column feed (Fcol1). These two loops take-up control as and
when the controller output becomes implementable (i.e. V2SP < V2MAX and L1SP < L1MAX).

Figure 15.2. Ethyl benzene process economic plantwide control structure (with long inventory loop)

It is highlighted that in the revised pairings for more conventional inventory control (Step
4 in Table 15.1), B2 must be controlled (by adjusting FC2/FTotBz SP) and not allowed to float as it
can result in a snowballing problem. This is because V2MAX is an active constraint at maximum
throughput implying limited capacity to boil-off EB in the product column. Any EB that could
not be boiled off in the product column would necessarily drop down the bottoms causing the
DEB recycle rate (B2) to slowly increase. To prevent this slow drift (snowballing), it must be
ensured that only as much EB is produced in the reaction section as can be boiled off in the
product column. This gets accomplished by adjusting the FC2/FTotBz SP to maintain B2, which
ensures the fresh ethylene feed to the process matches the EB boil-off rate. A seemingly
innocuous recommendation of allowing a self-optimizing CV to float and accepting the
consequent economic loss results in a very severe consequence of potential process instability.
This highlights the importance of Down's drill in ensuring the recommended control structure

159
does not suffer from such hidden instabilities due to slow accumulation of component
inventories.

Figure 15.3. Modified economic plantwide control structure for ethyl benzene process

If a conventional control system was designed for process operation around the design
condition, V2 would get used for maintaining a product column stripping temperature. As long as
the loop is functioning, the EB would get boiled-off and not accumulate in the DEB recycle loop.
However, once V2MAX goes active, product column stripping temperature control would be lost.
To ensure that the process does not succumb to snowballing in the DEB recycle loop, one would
have to design an override scheme that alters the material balance structure all the way up to the
process feed resulting in an inherently complicated scheme for constraint handling. In contrast,
the synthesized control structure is much simpler with no overrides and appealing in that the way
inventory is regulated remains the same regardless of the operating region.
Rigorous dynamic simulations are performed to test the synthesized control structure in
in Aspen Plus. All flow / pressure PI controllers are tuned tight for a fast and snappy servo
response, unless specified otherwise. The long B2 loop is tuned by hit-and-trial for a smooth
overall plantwide response. The non-reactive level controllers are P-only with a gain of 2. The
CSTR levels are controlled using a PI controller for offset free level tracking. The relay feedback
test feature with Tyreus-Luyben settings is used to obtain the CSTR level controller tuning
parameters at maximum throughput. All temperature measurements are lagged by 2 mins to
account for sensor and cooling / heating circuit dynamics. To tune the temperature loops, the

160
open loop step response at maximum throughput is obtained and the reset time set to 1/3rd of the
approximate 95% response completion time. The gain is then adjusted for a slightly
underdamped servo response with mild oscillations. The composition controllers are similarly
tuned. A sampling time and delay time of 5 mins each is applied to all composition
measurements. The tuning parameters of salient loops are reported in Table 15.2.
The closed loop dynamic response of the synthesized plantwide control system to a
throughput transition from the design throughput (FC2 = 630 kmol/h) to maximum throughput
(FC2 = 970 kmol/h) is shown in Figure 15.4. The product impurity is tightly controlled and the
transients in the process variables are smooth implying the suitability of the control structure for
near optimal operation over the wide throughput range.

Table 15.2. Salient Controller tuning parameter for Ethyl Benzene process
Controlled
KC τi (min) Sensor Span
Variable
LVLrxr1 5 250 0 – 100%
LVLrxr2 5 250 0 – 100%
Trxr1 4.8 25 0 – 400°C
Tcol1 3.2 18.5 77 °C – 157 °C
Tcol2 2 11 0 .0 – 244.7 °C
xBzD2 0.3 100 0 – 0.0016
xDEBD2 0.8 88.5 0 0.002
B2 0.2 1200 0 – 500 kmol/h
All level loops use KC = 2 unless otherwise specified
Pressure/flow controllers tuned for tight control
All composition measurements use a deadtime of 5 minutes and a sampling time of 5 mins

161
Figure 15.4. Low to maximum throughput transition of ethyl benzene process using
modified economic plant-wide control structure

162
 MODULE V
Chapter 16. Comprehensive Case Study I: Cumene Process
16.1. Process Description

. Figure 16.1 provides a schematic of the cumene process along with the design and base-case salient
operating conditions. Fresh benzene (C6) and fresh propylene (0.95 propylene and
0.05 propane), mixed with recycle benzene are vaporized in a vaporizer. The vapor stream is preheated using
the hot reactor effluent in a feed effluent heat exchanger (FEHE) before being heated to the reaction
temperature in a furnace. The heated stream is fed into a packed bed reactor (PBR), a shell and tube heat
exchanger with catalyst loaded tubes and pressurized coolant on the shell side. Propylene (C3) and C6 react in
the vapor phase to produce cumene (C9), which can further react with C3 to produce a small amount of
di-isopropyl benzene (C12 or DIPB) side product. The reactor effluent loses sensible heat in the FEHE and is
partially condensed in a cooler. The cooled stream with C9, C12, unreacted reactants and inert propane is fed to
a three column light-out-first distillation train. The purge column recovers inert propane and any unreacted
propylene with some benzene as vapor distillate. The bottoms is sent to the recycle column which recovers the
unreacted benzene as the distillate and recycles it. The recycle column bottoms is sent to the product column,
which recovers nearly pure C9 distillate and heavy C12 (+ some C9) bottoms

Figure 16.1. Cumene process schematic with salient design and base-case operating
 The reaction chemistry and kinetics used to model the process are provided in Table 16.1. The NRTL
physical property method is used to model thermodynamic properties. Steady state simulation was performed
19
using UniSim Design R390 version 3.61.0.0 from Honeywell. Luyben has studied the design and basic
regulatory control of a very similar cumene process flowsheet with the same reaction kinetics. The flowsheet
studied here differs in that the first distillation column replaces a flash tank to mitigate loss of precious benzene
in the C3 fuel gas stream. The optimized base-case process design and steady state operating conditions are also
6
shown in Figure 16.1. This revised design gives 6.8% higher profit than Luyben's flowsheet.

Table 16.1. Reaction chemistry and kinetics

Ea i Concentration
i Reaction ki
(kJ/kmol) terms fi(Cj)
1 C6H6 + C3H6 -C9H12 2.8 x 107 104174 CC3CC6
2 C9H12 + C3H6 -C12H18 2.32 x 109 146742 CC3CC9
Objective Maximize(J) J: hourly operating profit *
i
Reaction rate ri = ki.exp(-Ea /RT).fi(Cj)
-3 -3 -1
Cj in kmol.m; ri in kmol.m.s

16.2. Economic Plantwide Control System (CS1) Design

16.2.1. Step 0: Optimal Steady State Process Operation

The plant has a total of 12 steady state operating degrees of freedom (DOFs): 1 each for the two fresh
feeds, 1 for the furnace, 1 for reactor cooling, 1 for reactor pressure, 1 for reactor effluent cooler and 2 each for
the three distillation columns. Specification variables corresponding to these degrees of freedom chosen for
robust flowsheet convergence are: fresh
Total
propylene feed (FC3), total benzene flow (FC6 ), reactor inlet temperature (Trxr), reactor coolant
temperature (TRxrShell), reactor pressure (PRxr), reactor effluent cooler outlet temperature (Tcooler), D1 B1
first column vent temperature and bottoms propane mole fraction (Tvent and xC3 ), the recycle D2 B2
column distillate cumene and the bottoms benzene mole fractions (xC9 and xC6 ) and finally, D3 B3
the product column distillate cumene and the bottoms cumene mole fractions (xC9 and xC9 ). These 12
specification variables can be adjusted to achieve a given objective such as maximum throughput/profit or
maximum yield/selectivity.
In this work, the steady state hourly operating profit, P, defined as
P = [Product Revenue – Raw Material Cost – Energy Cost] per hour is used as a
quantitative economic criterion that is maximized using the available steady state DOFs. We consider two
modes of steady process operation. In Mode I, the desired throughput (production rate or feed processing rate)
is specified, usually based on business considerations. For processes with undesirable side products, such as the
cumene process considered here, the optimization typically attempts to maximize the yield to desired product.
For processes with no undesirable side products (e.g. a separation train), the optimization attempts to minimize
the energy consumption per kg product. In Mode II, the throughput itself is a decision variable for maximizing
the economic criterion. Often, the Mode II solution corresponds to steady process operation at/near the
maximum achievable throughput.

For the cumene process considered here, in Mode I, since the fresh propylene feed (FC3) is fixed, only the
remaining 11 DOFs need to be optimized. In Mode II, all 12 DOFs (including
FC3) need to be optimized. The optimization is subject to physical and operational process constraints such as
maximum / minimum material / energy flows, temperatures, pressures, product impurities etc.

Ideally all decision variables should be optimized simultaneously but this can result in an unwieldy
problem with poor convergence. The optimization is therefore simplified by applying engineering reasoning to
optimize only the dominant decision variables affecting the economic criterion with reasonable values for the
remaining decision variables. For the cumene process, the reactor effluent cooler temperature (Tcooler) has very
°
little impact on the economic objective function (P) and is therefore kept fixed at 100 C, a reasonable value
that ensures the reactor effluent vapor is condensed. Similarly, the yearly operating profit is insensitive to
changes around the base design values of the propane mol fraction leaking down the first column bottoms (xC3
B1 D2
) and the cumene mole fraction leaking up the second column distillate (xC9 ). These are therefore kept fixed
D1
at the base values. Also, the first column vapor vent stream temperature (Tvent ) is set by the cooling water at
32 °C.

These simple engineering arguments fix 4 specifications simplifying the optimization to 7 decision
variables for Mode I (given FC3) and 8 for Mode II. The optimization is performed using Matlab's fmincon
routine with Unisim as the back-ground steady state flowsheet solver. The constrained optimization problem
formulation (including price data and process constraints) and results for Mode I and Mode II are briefly
summarized in Table 16.2.
The optimization results are interpreted as follows. The minimum product purity
D3 MIN
constraint (xC9 = 99.9%) is active in both Mode I and Mode II, i.e. at all throughputs, for on-aim product quality
with no product give-away. The maximum reactor operating pressure
MAXMAX
(PRxr ) and maximum recycle (second) column boilup (V2 ) constraints are active at all throughputs. Reactor operation
at maximum operating pressure causes the reactor temperature to be lower for a given conversion improving
selectivity (cumene product yield). Recycle column operation at maximum boilup causes the total (fresh + recycle)
benzene to the reactor to be as high as possible, again enhancing selectivity with a higher reactor benzene to
MAX

propylene ratio. As throughput is increased, the product column maximum boilup constraint, V3 , goes active.
D2

Even as the throughput may be further increased by e.g. reducing the recycle column reflux (i.e. xC9 is
MIN

increased) and adjusting TRxr and TRxrShell to maintain conversion and selectivity, the Qfur constraint goes active
after which the selectivity decreases dramatically. The increase in throughput achieved is very marginal at < 1
MAX

kmol/h. We therefore treat V3 going active as corresponding to the maximum economic throughput (Mode II)
with FC3 = 169.96 kmol/h.

D3 MIN MAX MAX

The three Mode I active constraints (xC9 , PRxr and V2 ) along with the throughput specification (FC3)
leave four unconstrained DOFs. In Mode II, the throughput is not specified and gets determined by the value of
MAX
the additional V3 constraint so that the number of unconstrained DOFs remains four. The unconstrained
optimum values of the four decision
B3 B2
variables, xC9 , xC6 , TRxr and TRxrShell areB3reported in Table 16.2 for Mode I and Mode II.
The low Mode I optimum xC9 reduces the loss of precious cumene down the product column bottoms
B3
without a prohibitively high energy cost. The optimum Mode II xC9 is much higher at 10%. This reduces the
MAX
recycle column stripping load so that the V3 constraint goes active at higher throughputs for increased profit.
B3
Further loosening xC9 however causes the profit to decrease due to excessive cumene loss in the side product
stream. B2
The Mode I optimum benzene leakage down the recycle column bottoms, xC6 , is on the higher side at 0.09%
so that benzene is the principal cumene product impurity. This is reasonable as benzene is the cheaper product
impurity with DIPB consuming 2 extra mols of propylene. The
 B2
Mode II optimum xC6 value reduces to 0.05% so that the two product impurities are MAX
comparable. As shown in Figure 16.2, this balances throughput and selectivity with V2 and MAX B2
V3 active constraints. If xC6 is too high, the DIPB leakage in the product column distillate is prohibitively small
MAX
requiring high reflux so that the V3 constraint goes active at a significantly lower throughput. Similarly, if xC6
B2
is too low, the feed that can be processed by the recycle column maintaining its two separation specifications
MAX
without violating the V2 constraint is
B2 MAX
lower implying a loss in throughput. Also, as xC6 is loosened, with V2 active, the benzene recycle increases for
better selectivity with lower DIPB formation. Comparable amounts of the two principal impurities in the
product balances these two effects.

Table 16.2. Process optimization formulation and results' summary

Ea i Concentration
i Reaction ki
(kJ/kmol) terms fi(Cj)
1 C6H6 + C3H6 -C9H12 2.8 x 107 104174 CC3CC6
2 C9H12 + C3H6 -C12H18 2.32 x 109 146742 CC3CC9
Objective Maximize(J) J: hourly operating profit *
0 ≤ Material Flows ≤ 2 (base case) 0 ≤ V1, V2, V3 ≤ 1.5 (base case) Vent
Process Constraints Temperature = 32 oC 0 ≤ Energy Flows ≤ 1.7 (base case) 1 bar ≤ PRxr ≤
25 bar Cumene Product Purity ≥ 0.999 mol fraction
Decision Variable Mode I Mode II
FC3 101.93 kmol/h Fixed 169.96 kmol/h
FC6 Total 294.16 kmol/h 316.2 kmol/h
Trxr 322.26 °C 318.58 °C
TRxrShell 368.95 °C 367.98 °C
PRxr 25 bar Max 25 bar Max
Tcooler 100 °C Fixed 100 °C Fixed
Tvent D1 32 °C Fixed 32 °C Fixed
xC3 B1 0.1 % Fixed 0.1 % Fixed
xC9 D2 0.4 % Fixed 0.4 % Fixed
xC6 B2 0.09 % 0.05 %
xC9 D3 99.9 % Min 99.9 % Min
xC9 B3 0.4 % 10 %
Optimum J FC9 Active $3.809x103 h -1 93.59 kmol/h $5.879x103 h -1 150.045 kmol/h xC9 D3 MIN ,
Constraints xC9 D3 MIN , PRxr MAX , V2 MAX PRxr MAX , V2 MAX, V3 MAX
SNo CV Remarks on regulatory / economic significance
Determines process throughput.
1 FC3 Maximum throughput limited by V3 MAX
2 FC6 Total Increasing FC6Tot improves selectivity. Maximum FC6Tot limited by V2 MAX .
Affects reactor conversion and selectivity.
3 Trxr Stabilizes reaction heat recycle through FEHE. Affects reactor conversion
and selectivity. -1-1-1-1 1
*: Heater duty $16.8 GJ; Steam $9.83 GJ;Cooling water $0.16 GJ;FC6 $ 68.5kmol; FC3 $ 34.3kmol ; FC9 $
4 TRxrShell
-1 Stabilizes reaction heat removal. Operate at PRxr MAX for maximum reactor
150.0kmol
conversion.
5 PRxr Stabilizes gas inventory in reaction section.
We now seek a simple steadyDetermines
state operating policy
benzene that ensures
impurity level in near optimalFixed
product operation over the
by benzene entire
dropping
throughput range. For economically optimal operation, we would like tight control of the active constraints and
appropriate management of the remaining unconstrained steady state DOFs using SOVs. Preferably, the CVs
corresponding to the unconstrained steady state DOFs should be measurements that are cheap, reliable, fast,
robust and dynamically well
behaved with respect to the manipulated variables (MVs). These CVs should therefore be flow, pressure
and temperature based avoiding cumbersome analytical measurements.

Figure 16.2. Optimum benzene impurity level in cumene product

B1
Of the 12 decision variables in Table 16.2, 4 decision variables, Tvent, TCooler, xC3 and
D2 D3 MIN xC9 were fixed at reasonable values. In Mode I, there are three active constraints, xC9 , MAX MAX MAX

PRxr and V2 along with a specified FC3. In Mode II, V3 going active sets FC3. Optimum values for the remaining
D3 B3
4 unconstrained decision variables in both modes, TRxr, TRxrShell, xC6 and xC9 were obtained.
B1
In the above set of variables, compositions not related to the product quality, i.e., xC3 ,
D2 B3 xC9 and xC9 would usually not be available. Accordingly, we consider using appropriate temperature
inferential measurements. On the purge and product columns, controlling appropriate sensitive stripping tray
S S
temperatures, T Col1 and T Col3, respectively, would regulate
B1 B3
the light key leakage down the bottoms. This would indirectly maintain xC3 and xC9 within a small band. On the
recycle column, maintaining the reflux (L2) in ratio with the column feed (B1) would regulate the distillate
D2 D3
cumene leakage (xC9 ). The product DIPB impurity mol fraction (xC12 ) and benzene impurity mol fraction
D3
(xC6 ) measurements would usually be available in
D3 D3MIN D3
an industrial setting. For on-aim product cumene mol fraction (xC9 = xC9 = 99.9%), xC6 D3 D3
+ xC12 = 0.1% so that only one of the impurity mol fractions is independent. We take xC6 to D3 D3
be independent with xC12 = 0.1% -xC6 .
The revised practical CVs corresponding to the 12 steady state DOFs are tabulated in Table 16.3 along with
their regulatory and economic significance. The CVs are the active constraints (or specifications) and four
D3 S
unconstrained CVs, TRxr, TRxrShell, xC6 and T Col3. Of the unconstrained CVs, the optimum reactor inlet
temperature (TRxr) and reactor coolant temperature
(TRxrShell) are nearly the same for Mode I and Mode II (see Table 16.2). Holding these two variables constant
would likely be near optimal across the wide throughput range. For the remaining two CVs, since economic
losses per unit deviation away from the optimum values are usually the highest at maximum throughput, we
consider implementing the Mode II optimum value at the lower throughputs. This gives a very simple constant
setpoint policy across the entire throughput range. To quantify the economic loss entailed, Figure 16.3
compares the variation with throughput in the optimum operating profit and the operating profit using the
constant Mode II setpoints for the above four CVs. The constant setpoint operating policy provides near
optimal steady operation with the maximum profit loss being < 0.21%. These four CVs may thus be deemed as
SOVs that provide near optimum steady operation across the entire throughput range.

Table 16.3. Revised practical CVs

Ea i Concentration
i Reaction ki
(kJ/kmol) terms fi(Cj)
1 C6H6 + C3H6 -C9H12 2.8 x 107 104174 CC3CC6
2 C9H12 + C3H6 -C12H18 2.32 x 109 146742 CC3CC9
Objective Maximize(J) J: hourly operating profit *
0 ≤ Material Flows ≤ 2 (base case) 0 ≤ V1, V2, V3 ≤ 1.5 (base case) Vent
Process Constraints Temperature = 32 oC 0 ≤ Energy Flows ≤ 1.7 (base case) 1 bar ≤ PRxr ≤
25 bar Cumene Product Purity ≥ 0.999 mol fraction
Decision Variable Mode I Mode II
FC3 101.93 kmol/h Fixed 169.96 kmol/h
FC6 Total 294.16 kmol/h 316.2 kmol/h
Trxr 322.26 °C 318.58 °C
TRxrShell 368.95 °C 367.98 °C
PRxr 25 bar Max 25 bar Max
Tcooler 100 °C Fixed 100 °C Fixed
Tvent D1 32 °C Fixed 32 °C Fixed
xC3 B1 0.1 % Fixed 0.1 % Fixed
xC9 D2 0.4 % Fixed 0.4 % Fixed
xC6 B2 0.09 % 0.05 %
xC9 D3 99.9 % Min 99.9 % Min
xC9 B3 0.4 % 10 %
Optimum J FC9 Active $3.809x103 h -1 93.59 kmol/h $5.879x103 h -1 150.045 kmol/h xC9 D3 MIN ,
Constraints xC9 D3 MIN , PRxr MAX , V2 MAX PRxr MAX , V2 MAX, V3 MAX
SNo CV Remarks on regulatory / economic significance
Determines process throughput.
1 FC3 Maximum throughput limited by V3 MAX
2 FC6 Total Increasing FC6Tot improves selectivity. Maximum FC6Tot limited by V2 MAX .
Affects reactor conversion and selectivity.
3 Trxr Stabilizes reaction heat recycle through FEHE. Affects reactor conversion
and selectivity.
4 TRxrShell Stabilizes reaction heat removal. Operate at PRxr MAX for maximum reactor
conversion.
16.2.2.
5 Step 1: Loops
PRxr for Tight Economic
Stabilizes CV Control
gas inventory in reaction section.
Determines benzene impurity level in product. Fixed by benzene dropping
6 xC6 D3 MAX MAX
down theatrecycle
The hard active constraints maximum column.
throughput are V2 and V3 . These are economically important
MAX
as a back-off from V2 reduces the benzene recycle rate with loss in reactor selectivity while a back-off in
MAX
V3 causes a loss in throughput. To minimize the back-off, V2 and V3 are controlled tightly using the respective
reboiler duties (QReb2 and QReb3).
 MAX
PRxr , another economically important active constraint due to its impact on the reactor conversion, is
considered a soft constraint. The reactor pressure is controlled tightly around its
SP MAX
maximum value (PRxr = PRxr ) by manipulating the pressure regulatory valve (PRV) between the reaction and
separation sections. The pairing would provide tight control.

Figure 16.3. Comparison of optimum steady profit and achieved profit using simple
constant setpoint operating policy at various throughputs

Economic operation requires tight control of the product impurity levels for on aim
D3 MIND3
product purity of xC9 , a soft active constraint. For maintaining xC9 , the two principal impurities in the product, C12
D3

and C6, must be maintained. Control of xC12 is accomplished by adjusting the product column reflux to feed ratio
D3

(L3/B2). The ratio scheme helps mitigate the variability in xC12 due to the feedforward action of the ratio controller
to column feed flow disturbances. With regard to the C6 impurity in the product, note that all the benzene that
leaks down the recycle column ends up in the product. Tight regulation of the benzene leakage down the recycle
S MAX

column can be achieved by maintaining a stripping tray temperature (T Col2). Since V2 constraint is active, we
may use the feed to the recycle column (B1) or the recycle column reflux rate (L2) as the MV. The former would be
effective for a mostly liquid feed and the latter must be used for a mostly vapor feed. For the specific choice of the
S

design pressures of the purge and recycle columns, the B1 vapor fraction is ~25% so that the T Col2 -B1 pairing is
selected. The

S SP D3
TCol2 is adjusted by a xC6 composition controller. The product impurity mol fraction
 D3SP D3SP D3SP
setpoints are chosen as xC6 = 0.05% (Mode II optimum value) and xC12 = 0.1% -xC6 = 0.05%. These setpoints
are held constant at lower throughputs for near optimal operation. Economic operation requires the cumene
leakage down the product column bottoms to be small. This is achieved by maintaining a product column
S
stripping tray temperature (T Col3). MAX S
Since V3 is active and the column feed (B2) is mostly liquid, the TCol3 -B2 pairing is chosen.
Lastly, maintaining a high reactor conversion for a small propylene loss in the fuel gas stream as well
as a high reactor selectivity for small loss of precious raw materials as DIPB by-product are economically
important objectives. Holding the reactor inlet temperature constant at 322 °C and the reactor shell side coolant
temperature at 367 °C ensure that the reactor conversion and selectivity are maintained at high values across the
entire throughput range. TRxr is controlled tightly by manipulating the furnace duty (Qfur) for tight control.
TRxrShell = 367 °C is a direct input (MV) to the process as the constant coolant temperature model is used in the
simulations. In practice, since the reactor temperature is high, a proprietary heating oil such as Dowtherm
would be used as the coolant with high pressure steam being generated in a downstream Dowtherm heated
boiler. TRxrShell then is controlled by adjusting the boiler pressure setpoint with the boiler pressure being
controlled by the exit steam flow.

16.2.3. Step 2: Inventory Control System Design

We now pair loops for inventory regulation, inventory being interpreted in its most comprehensive
sense to include total material, phase, components and energy. Of the 12 steady state DOFs, 8 loops have
already been implemented in Step 1. This leaves 4 additional loops that need to be configured plus loops for
regulating the reflux drum and bottom sump levels on the three columns along with the column pressures and
the feed vaporizer level.
S
The 4 additional loops correspond to holding L2/B1, Tvent, T Col1 and TCooler at their design values.
Maintaining L2/B1 using a feed to reflux ratio controller regulates the C9 leakage in the benzene recycle stream.
The purge column condenser temperature is controlled by manipulating its condenser duty (QCnd1). This
regulates the loss of precious benzene in the fuel gas stream. The purge column stripping tray temperature
S
(T Col1) is controlled using its boilup (V1) to regulate the C3 leakage down the bottoms. The reactor effluent
condensate temperature (TCooler) is controlled by manipulating the effluent cooler duty (QCooler). This ensures
proper regulation of the gas/vapor inventory in the reaction section in conjunction with the PRxr control loop.

The recycle and product column pressures (PCnd1 and PCnd2) are regulated by the respective condenser
duty valves, QCnd2 and QCnd2. The purge column pressure (PCol1) is regulated by the vent rate, D1. Its reflux
drum level (LVLRD1) is regulated by manipulating the reflux (L1). The feed vaporizer level (LVLVap) is regulated
by the vaporizer duty (QVap). The recycle column and product column reflux drum levels (LVLRD2 and LVLRD3)
are regulated using the respective distillate rates (D2 and D3). The product column bottom sump level (LVLBot3)
is regulated using its bottoms rate (B3). With these pairings, no close-by valves are left for regulating the purge
column and recycle column bottom sump levels (LVLBot1 and LVLBot2). The only option is to manipulate the two
fresh feeds, FC3 and FC6. C3 is the limiting reactant with near complete single-pass conversion so that FC3
determines the cumene and DIPB production in the reactor. Since the cumene and DIPB accumulate at the
bottom of the recycle column, the LVLBot2 -FC3 pairing is implemented for recycle column sump level control
with the LVLBot1 -FC6 pairing being implemented for purge column sump level control.
16.2.4. Step 3: Throughput Manipulation and Additional Economic Loops
MAX
In this example, there is only one active constraint region corresponding to V3 going active at
maximum throughput with the other constraints / specifications being fixed at their Mode II values at lower
throughputs. The throughput may be reduced by reducing V3 below
MAX SP
V3 . V3 is then the throughput manipulator (TPM) adjusted to operate the plant at the desired throughput below
maximum. There are no additional SOVs whose control needs to be taken up at lower throughputs as no
additional constraints become inactive at lower throughputs.
The economic plantwide control structure, labeled CS1, obtained by the application of Step 1-3 is
shown in Figure 16.4 with the economic loops in blue. CS1 has been designed for the tightest possible control
of the economic CVs using close by MVs. Since control valves get used up in these loops, in the inventory
control system, the MVs of the bottom sump level loops for the purge and recycle columns are not local to the
respective units but away at the fresh feeds and thus very unconventional. Even so, acceptable level regulation
is expected as the lag associated with the reaction section is small with the material essentially flowing through
a long pipe with small vaporizer and the reactor effluent cooler lags. The acceptable level regulation and overall
process stabilization was confirmed from rigorous dynamic simulations. With the unconventional long level
loops, the control structure attempts tight control of the economic CVs with loose level control. In other words,
the structure attempts tight control of the economic CVs by transforming the transients to the surge levels that
have no steady state economic impact.

Figure 16.4. Economic plantwide control structure (CS1)

16.3. Conventional Plantwide Control Structure (CS2)

The conventional plantwide control structure, CS2, with the TPM at the C3 (limiting reactant) feed is shown
in Figure 16.5. The total benzene (fresh + recycle) is maintained by FC6Tot to prevent snowballing in the benzene
recycle loop. In the reaction section, LVLVap is controlled by QVap, TRxr is controlled by QFur, TRxrShell is set at its near
MAX

optimum value, PRxr is controlled at PRxr by the PRV and the partially condensed reactant effluent temperature
(TCooler) is maintained by its cooling duty, QCooler. In the separation train, the recycle and product column pressures
are controlled by the respective condenser duties, the reflux drum levels using the respective distillate streams and
the bottom sump levels using the respective bottoms streams. On the purge column, the column pressure is controlled
by the vapor vent, the overhead condenser temperature is maintained by the condenser duty and the reflux drum level
S

is controlled by the reflux. To regulate the C3 leakage down the bottoms, T Col1 is maintained by V1. On the recycle
S

column, L2 is maintained in ratio with the column feed (B1) and T Col2 is maintained by V2 with

S SP D3
TCol2 being adjusted to maintain the product impurity xC6 . On the product column, the reflux (L3) is maintained
SP D3 S
in ratio with the feed (B2) and L3/B2 is adjusted to maintain the product impurity xC12 . T Col3 is maintained
by adjusting V3.

Figure 16.5. Conventional plantwide control structure, CS2, with overrides

MAX
Since optimal operation requires running the process at V2 at all throughputs, a supervisory controller is
SP
installed that adjusts the total benzene setpoint (FC6Tot ) to maintain V2
 MAX
at its setpoint. Since V2 is a hard constraint corresponding to the initiation of recycle column flooding and
S
since control of the stripping tray temperature (T Col2) must never be lost to ensure the product benzene
MAX
impurity level is always regulated, some back-off from the V2 limit would be needed to ensure the hard
constraint is not violated during worst case transients. MAX
The other hard constraint that must be handled is V3 , the bottleneck constraint, which goes active as
MAX
throughput is increased towards maximum. When V3 goes active, product column temperature control
S
(T Col3) is lost implying loss of precious cumene down the bottoms with a severe economic penalty. To avoid
the same, an override control system is put in place that alters the material balance control structure all the way
MAX
up to the C3 feed to ensure that column temperature control is not lost when V3 goes active, as in Figure 16.5.

The override scheme works as follows. The override temperature controller on the product column is
S MAX
direct acting and has its setpoint slightly below the T Col3-V3 loop setpoint. Thus when V3 is inactive, its
MAX
output is high (usually saturated) and B2 controls the recycle column sump level. When V3 goes active,
product column temperature decreases below the second temperature controller setpoint and its output
ultimately decreases below the LVLBot2 controller output with the low select passing the manipulation of B2
from the LVLBot2 controller to the override temperature controller. Once this occurs, LVLBot2 control is lost and
it rises. The second LVLBot2 override controller then takes over manipulation of B1 via the low select in a
manner similar to the product column temperature override scheme. This causes LVLBot1 control to be lost and
the second LVLBot1 override controller ultimately takes over FC3 manipulation. The override scheme thus works
MAX
to cut down on the fresh propylene feed on V3 going active.

16.4. Dynamic Simulation Results and Discussion

Rigorous dynamic simulations are performed in Unisim to evaluate and compare the performance of
the synthesized economic plantwide control structure, CS1, with the conventional plantwide control structure,
CS2.

16.4.1. Controller Tuning

A consistent procedure is used to tune the various controllers. All flow and pressure controllers are PI
and tuned for a fast and snappy response. All conventional level controllers with local unit specific pairings are
P only and use a gain of 2 to smooth out flow transients. The temperature controllers are PI with a 45 s sensor
lag. The Unisim autotuner is used to obtain a reasonable value of the reset time and controller gain (KC). The
KC is then adjusted for a fast but not-too-oscillatory servo response. All composition controllers use a sensor
dead-time and sampling time of 5 mins. The autotuner does not provide reasonable initial tuning parameters so
rd
that the open loop response is first obtained and the reset time set to 2/3 open loop response completion time
and KC set to the inverse of the process gain. These tunings work well for the two product impurity controllers
in both CS1 and CS2.

In CS1, the unconventional non-local LVLBot1 and LVLBot2 controllers are P only and are tuned initially by hit
and trial to stabilize the process. The temperature and composition loops are then tuned as discussed above.
Finally, the non-local level controller tunings are further refined for a smooth overall plantwide response to the
principal disturbances. In CS2, the product column override temperature controller setpoint is chosen to the
highest possible value so that the over-ride controller never goes active for the different disturbance scenarios.
This gives a
setpoint that is 2 °C below nominal. The LVLBot1 and LVLBot2 override setpoints are chosen 10% above the
MAX
nominal setpoint of 50%. Also, aggressive tuning is attempted to ensure FC3 is cut quickly when V3 goes
active to mitigate the loss of precious cumene down the product column bottoms during the transient. Both the
over ride level controllers are P only. Finally the supervisory recycle column boilup controller is tuned for a
not-too-oscillatory servo response. The salient controller tuning parameters and setpoints thus obtained are
reported in Table 16.4 for CS1 and CS2.

*#
Table 16.4. CS1 and CS2 controller parameters
Ea i Concentration
i Reaction ki
(kJ/kmol) terms fi(Cj)
1 C6H6 + C3H6 -C9H12 2.8 x 107 104174 CC3CC6
2 C9H12 + C3H6 -C12H18 2.32 x 109 146742 CC3CC9
Objective Maximize(J) J: hourly operating profit *
0 ≤ Material Flows ≤ 2 (base case) 0 ≤ V1, V2, V3 ≤ 1.5 (base case) Vent
Process Constraints Temperature = 32 oC 0 ≤ Energy Flows ≤ 1.7 (base case) 1 bar ≤ PRxr ≤
25 bar Cumene Product Purity ≥ 0.999 mol fraction
Decision Variable Mode I Mode II
FC3 101.93 kmol/h Fixed 169.96 kmol/h
FC6 Total 294.16 kmol/h 316.2 kmol/h
Trxr 322.26 °C 318.58 °C
TRxrShell 368.95 °C 367.98 °C
PRxr 25 bar Max 25 bar Max
Tcooler 100 °C Fixed 100 °C Fixed
Tvent D1 32 °C Fixed 32 °C Fixed
xC3 B1 0.1 % Fixed 0.1 % Fixed
xC9 D2 0.4 % Fixed 0.4 % Fixed
xC6 B2 0.09
*: All level % use KC = 2 unless otherwise specified
loops 0.05#:%
xC9 D3 99 9 % Mi controllers tuned for tight control
Pressure/flow 99 9 % Mi

16.4.2. Closed Loop Results

CS1 and CS2 are dynamically tested for different disturbance scenarios. First, the dynamic transition
from Mode I to Mode II is simulated. The dynamic response is also obtained for a ±10% throughput step
change and a ±3% step change in the feed propylene mol fraction for Mode I (FC3 = 101.93 kmol/h) operation.
For Mode II, the dynamic response is obtained for the latter as well as a ±5% step bias in the FC3 flow sensor.
MAX
For convenience, the CS2, supervisory V2 controller setpoint is set at V2 even as in practice sufficient back-off
MAX
would be provided to ensure the hard V2 constraint is never violated during worst case transients and benzene
impurity control in the product cumene is never lost.

We first consider throughput transition using CS1 and CS2, from Mode I (low throughput) to Mode II
(maximum throughput) and back. In both structures, the TPM is ramped at a rate that causes FC3 to change by
~10 kmol in 15 hrs. This ensures that the severity of the throughput transition disturbance is comparable in both
the structures. For the throughput
SPMAX
transition in CS1, V3 , is ramped up at a rate of 0.79 kmol/h to V3 , held constant for 20 hours and then ramped
SP
back down at the same rate. In CS2, FC3 is ramped at a rate of 0.74 kmol/h till 184 kmol/h (or lower if
override takes over FC3 manipulation), held there for about 30 hours to
allow for the over-rides to take over and stabilize and then ramped back down to 101.93 kmol/h. As
21 S
recommended by Shinskey , we use external reset on the PI T Col3 override controller to ensure it takes up
MAX
B2 manipulation at the earliest once V3 goes active.
The CS1 and CS2 transient response of salient process variables is plotted in Figure 16.6. Tight
product purity control as well as smooth plantwide transients are observed for both CS1 and CS2. In CS2, the
MAX
major events of V3 going active (P1), the ethylene feed being cut by the LVLBot1 override (P2) and beginning
SP
of the FC3 (TPM) ramp down (P3) are shown. In the CS2 dynamic response, oscillations post LVLBot1 override
controller taking over FC3 manipulation are
MAX SP
seen. Also it takes about 5 hrs between V3 going active and FC3 manipulation passing to the B3 MAX
LVLBot1 override. The transient xC9 response for CS1 and CS2 also shows that once V3 goes active, the cumene
leakage in the DIPB stream remains well regulated in CS1 while in CS2 the leakage increases due to the lower
S
T pur override setpoint. In the entire transient period, LVLBot1 and LVLBot2 vary within a band of 15% and 24%
respectively, in CS1. The corresponding figures for CS2 are comparable at 16% and 24% respectively.

To compare the structures for Mode II operation, Figure 16.7 plots the dynamic response of important
process variables to a ±5% step bias in the FC3 measurement for CS1 and CS2. The dynamic response for CS1
achieves tight product purity control with a settling time of about 10 hours. Similarly, the CS2 transient
response also completes in about 10 hours. Note that since
MAX S
V3 is active, the CS2 TCol3, LVLBot2 and LVLBot1 overrides are on and the material balance control structure is
oriented in the reverse direction of process flow.
To compare the structures for Mode I operation, Figure 16.8 plots the plantwide dynamic response of
important process variables to a step change in the TPM for a ±10% throughput change. In CS1, to bring about
SP
a 10% increase and decrease in FC3, the V3 must be changed by +22.1 kmol/h and -21.9 kmol/h, respectively.
SP
In CS2, FC3 is directly set by FC3 (TPM). The product purity and DIPB cumene loss control in CS2 is not as
tight as in CS1 as the TPM for CS1 is located at the product column. In CS2, on the other hand, the TPM is at a
process feed and the downstream product column gets subjected to a less severe transient due to filtering by the
intermediate units. Overall, a smooth plantwide response is observed in both structures. The response
completion time for CS1 and CS2 is slightly above and below 10 hrs, respectively.

Figure 16.9 compares the plantwide response of important process variables to a ±3% step change in
the C3 feed propane (inert) impurity in Mode I operation. Both structures handle the disturbance well with the
product purity being tightly controlled. The overall plantwide response is also smooth with a response settling
time of about 15 hrs for CS1 and about 10 hrs for CS2.
 (a)

(b)

Figure 16.6. Transient response for throughput transition. (a) CS1; (b) CS2
 (a)

(a) CS1; (b) CS2 ―: -5% bias;

···: +5% bias
 (a)

(b)
 (a)

(b)

Figure 16.8. Mode I transient response to ±10% throughput change. (a) CS1; (b) CS2 ―: -10%;
···: +10%
(a) CS1; (b) CS2
―: -3% ; ···: +3%
16.4.3. Quantitative Dynamic and Economic Comparison of CS1 and CS2

In this subsection, the dynamic and economic performance of CS1 and CS2 is quantitatively compared.
In addition to the disturbance scenarios already considered, we consider a -5% step bias in FC3 measurement
MAX
with the initial steady state corresponding to V3 -V3 approaching 0 (Mode II). The overrides in CS2 are then
'ready to be triggered'.
D3 B3
To quantify the dynamic performance, the IAE values for xC9 and xC9 for the 10h transient period post
disturbance are reported in Table 16.5. From the data, it is evident that both structures provide comparable
MAX
regulation of product purity and the cumene loss in the byproduct stream in Mode I (V3 inactive) for a feed
B3
propylene composition change. For a ramped throughput change, even as the regulation of xC9 is significantly
poorer in CS1, it is acceptably
B3 SP
small. As already noted, the larger xC9 variability in CS1 is because the CS1 TPM (V3 ) is located at the product
column. The Mode I throughput change data (row 1) also suggests that CS2 achieves slightly tighter product
purity control. For Mode II operation, the data (rows 3 and
D3 B3
4) suggests that CS1 and CS2 provide comparable dynamic regulation of xC9 and xC9 for process feed
disturbances, namely, a 3% step change in the propylene feed composition or a 5% step bias in the FC3 sensor.
B3 S
The IAE values for xC9 with the T Col3 override about to be triggered (last two rows) with and without
external reset suggest that Shinskey's simple external reset scheme significantly improves the tightness of
control by ensuring that the unselected output does not deviate too far away from the selected output due to
reset windup.

D3 B3
Table 16.5. IAE values for xC9 and xC9 for 10 h transient post disturbance
Ea i Concentration
i Reaction ki
(kJ/kmol) terms fi(Cj)
1 C6H6 + C3H6 -C9H12 2.8 x 107 104174 CC3CC6
2 C9H12 + C3H6 -C12H18 2.32 x 109 146742 CC3CC9
Objective Maximize(J) J: hourly operating profit *
0 ≤ Material Flows ≤ 2 (base case) 0 ≤ V1, V2, V3 ≤ 1.5 (base case) Vent
Process Constraints Temperature = 32 oC 0 ≤ Energy Flows ≤ 1.7 (base case) 1 bar ≤ PRxr ≤
25 bar Cumene Product Purity ≥ 0.999 mol fraction
Decision Variable Mode I Mode II
FC3 101.93 kmol/h Fixed 169.96 kmol/h
FC6 Total 294.16 kmol/h 316.2 kmol/h
Trxr 322.26 °C 318.58 °C
TRxrShell 368.95 °C 367.98 °C
PRxr 25 bar Max 25 bar Max
Tcooler 100 °C Fixed 100 °C Fixed
Tvent D1 32 °C Fixed 32 °C Fixed
xC3 B1 0.1 % Fixed 0.1 % Fixed
xC9 D2 *: Initial 0.4 % state
steady Fixed &: TPM setpoint ramped over 0.4
6 h. %
IAEFixed
calculated over
0 09 15
% h period #: CS2 overrides are ‘ready to be
0 05 %
S
triggered’ %: No external reset in CS2 T override
S
pur @:External reset in CS2 T pur
override

To quantify the economic performance, the Mode I and Mode II steady state hourly profit
SP MAX MAX
is reported in Table 16.6. In CS2, V2 is backed-off from V2 by the least amount for which the V2 constraint
does not get violated for the worst-case disturbance scenario, which is a -5% step bias in FC3, requiring the
MAX
maximum back-off from V2 . Negligible back-off is needed in
 MAXMAX
CS1 which is designed for process operation at V2 . Due to the back-off from V2 in CS2, its steady profit is
slightly lower (up to >0.1% in Mode II) than CS1.

Table 16.6. Steady state and transient profit data for CS1 and CS2
Steady state hourly profit data Ea i Concentration
i Reaction ki
(kJ/kmol) terms fi(Cj)
1 C6H6 + C3H6 -C9H12 2.8 x 107 104174 CC3CC6
2 C9H12 + C3H6 -C12H18 2.32 x 109 146742 CC3CC9

Av Av
Transient profit data (IEP and ΔIEP values)
Ea i Concentration
i Reaction ki
(kJ/kmol) terms fi(Cj)
1 C6H6 + C3H6 -C9H12 2.8 x 107 104174 CC3CC6
2 C9H12 + C3H6 -C12H18 2.32 x 109 146742 CC3CC9
Objective Maximize(J) J: hourly operating profit *
0 ≤ Material Flows ≤ 2 (base case) 0 ≤ V1, V2, V3 ≤ 1.5 (base case) Vent
Process Constraints Temperature = 32 oC 0 ≤ Energy Flows ≤ 1.7 (base case) 1 bar ≤ PRxr ≤
25 bar Cumene Product Purity ≥ 0.999 mol fraction
Decision Variable Mode I Mode II
FC3 101.93 kmol/h Fixed 169.96 kmol/h
FC6 Total 294.16 kmol/h 316.2 kmol/h
Trxr 322.26 °C 318.58 °C
TRxrShell 368.95 °C 367.98 °C
PRxr 25 bar Max 25 bar Max
Tcooler 100 °C Fixed 100 °C Fixed
Tvent D1 32 °C Fixed 32 °C Fixed
xC3 B1 0.1 % Fixed 0.1 % Fixed
xC9 D2 0.4 % Fixed 0.4 % Fixed
xC6 B2 0.09 % 0.05 %
xC9 D3 99.9 % Min 99.9 % Min
xC9 B3 *: Initial steady0.4
state
% &: TPM setpoint ramped over 6 h. IAE10 %
calculated over 15 h period #: CS2 overrides are ‘ready to be triggered’
Optimum J FC9 Active $3 809x103 h 1 93 59 kmol/h $5 879x103 h 1 150 045 kmol/h xC9 D3 MIN
S
%: No external reset in CS2 T override
S
pur @:External reset in CS2 T pur
override

To quantify economic losses during transients, Table 16.6 also reports the time average integral error
for the 10 hour transient period (T) post disturbance defined as

SS
where Pt is the instantaneous hourly profit and Pf is the final steady state hourly profit for a disturbance. The
metric is thus the time average cumulative transient profit deviation from the final steady state profit. Positive
(negative) values indicate the extra hourly profit (loss) over the final steady state profit in the transient period.
One would expect that any transient profit for a disturbance in one direction would be nullified by a similar
Av
transient loss for the same disturbance in the opposite direction. The IEP values for a given disturbance in
either direction should thus be approximately the same magnitude but opposite signs. A large negative
difference between the two corresponds to an unrecoverable transient economic loss. Table 16.6 also reports
this difference
 Av Av+ Av
ΔIEP = IEP -IEP
Av+ Av
where IEP and IEP correspond to an increase and decrease, respectively, in the disturbance magnitude. As
Av
expected, in all but one disturbance scenario, ΔIEP is small for both CS1 and CS2. For a ±5% step change in
Av
the FC3 measurement with the CS2 overrides 'ready-to-betriggered', the ΔIEP is large negative implying
significant unrecoverable transient losses. These losses are attributed to the excessive leakage of precious
MAX S
cumene in B3 between V3 going active and T Col3 override taking over B2 manipulation. Every extra mol of
lost cumene consumes expensive reactants that cost twice the product to raw material price difference.
S
Regardless of whether external reset is used or not on the T pur override, the transient profit loss is significant
at >4.5% of the steady state Mode II profit. The transient loss figures with and without external reset are
B3
comparable as the oscillatory xC9 response for the no external reset leads to cancellation of errors in the
undershoots and overshoots.
SP
If the CS2 overrides are switched off (e.g. by an operator), FC3 must be sufficiently reduced from the
MAX
maximum achievable throughput so that the V3 constraint does not get violated during the worst-case
transient, which is a -5% step change in the FC3 measurement. This back-off results in a significant steady
hourly profit loss of >4% due to lower maximum throughput. The results demonstrate that CS2 with overrides
or backed-off operation results in non-negligible economic loss.

16.4.4. Discussion

The results for the case study suggest that the economic plantwide control structure, CS1, designed for
tightest possible control of the economically important hard active constraints
MAX MAX
(V3 and V2 ), achieves superior economic process operation particularly in Mode II, compared to the
conventional control structure, CS2. CS1 is also simpler than CS2 in that the inventory management strategy
MAX
remains fixed regardless of whether the V3 constraint is active or not. CS2 on the other hand is more
complicated requiring 3 additional override controllers to alter the material balance control structure all the way
MAX
up to the C3 feed, once the V3 constraint goes active. Proper tuning and setpoint selection of these override
controllers is necessary to ensure that they get activated in the proper order without too much time elapsing
MAX
between when V3 goes active and the overrides 'take-over' control. Proper design of the override scheme can
be tricky and for severe enough transients, the correct override order may get violated and large plantwide
transients can occur due to the overrides 'taking-over' and 'giving-up' control, similar to 'on-off' control. One
such occurrence and operators would be inclined to turn the scheme off and resort to the more conservative
backed-off process operation with a significantly more severe economic penalty.

It is also worth noting that in our analysis, we have considered only a single disturbance
MAX MAX
to be active at a time and the hard maximum boilup constraints (V2 and V3 ) to be constant. In practice, multiple
disturbances are active all the time. More importantly, the hard maximum boil-up constraint limits themselves
are transient, depending on the feed flow and reflux flow as well as other factors such as impurities that
build-up over time inside the column. The CS2 economic performance is therefore likely to be significantly
MAX

inferior to CS1 due to the need for a higher back-off in V2 as well as unrecoverable transient cumene loss in
MAX

the DIPB stream with the override scheme switching on and off due to variability in the V3 limit.

SP
The major difference between CS1 and CS2 is in the location of the TPM; V3 for CS1
SP SP and FC3 for CS2. Since V3 is the last constraint to go active (i.e. the bottleneck constraint) and
also economically important with any back-off resulting in reduced throughput, it makes sense to use it as the
TPM and not for the conventional control task of tray temperature control. Typically, due to the high sensitivity
of recycle flows to throughput changes (snowball effect), the bottleneck constraint is usually inside the recycle
loop. The case study results support the heuristic of locating the TPM at the bottleneck constraint for economic
operation.
Lastly, we highlight that the conventional practice in control structure design is to implement inventory
control loops with their MVs being 'local' to the specific unit containing the inventory. The basic idea is to
ensure that the inventory loops are robust. This case study illustrates that it is possible to develop control
structures with seemingly unworkable 'long' inventory control loops that provide acceptable regulation with
tight control of the economic CVs over the entire throughput range. The top-down pairing philosophy, as
illustrated here should be applied to come up with such unconventional but workable economic plantwide
control structures, in the knowledge that should the inventory control be fragile, the pairings can always be
revised towards 'local' inventory loops and 'long' economic loops in lieu.

16.5. Conclusions

In conclusion, this article demonstrates through a case-study, the crucial role of economically
important maximum throughput hard active constraints in determining the input-output pairings for economic
plantwide control. The approach demonstrated here leads to a simple control structure with unconventional
inventory loops for process operation over the entire throughput range. Conventional control systems that do
not take into consideration the active constraints on the other hand must resort to complicated overrides for
constraint handling at high throughputs, with overall inferior economic performance.
 Chapter 17. C4 Isomerization Process

17.1. Process Description

Figure 17.1 shows a schematic of the C4 isomerization process studied in this work. A fresh C4 stream
containing n-C4 and i-C4 with some C3 and i-C5 impurities is fed to a deisobutanizer (DIB) column that recovers
i-C4 with some n-C4 (heavy key) impurity as the distillate. All the C3 in the fresh C4 feed leaves in the distillate.
The DIB bottoms consisting of nC4, i-C5 and some i-C4 (light key) impurity is fed to a purge column that
recovers i-C5 with some n-C4 (light key) as the bottoms. The purge column distillate consisting of C4’s and
some i-C5 (heavy key) is fed to an adiabatic packed bed reactor (PBR) after preheating in a feed effluent heat
exchanger (FEHE) followed by heating to the reaction temperature in a heater. The n-C4 isomerizes in the PBR
to i-C4. The hot reactor effluent preheats the cold reactor feed in the FEHE and is then condensed in a flooded
cooler. The subcooled liquid is rich in i-C4 and is fed to the DIB column above the relatively i-C4 lean fresh C4
17
feed. The base-case process design and steady state operating conditions (adapted from Luyben et al. ) are
shown in Figure 17.1. The irreversible reaction kinetic model in their work is used along with the SRK equation
of state to model the thermodynamic properties. Aspen Hysys is used for steady state and dynamic process
simulation. Hysys uses the sequential approach for steady state solution of flowsheets with Wegstein updation
at the recycle tear. The inside-outside algorithm is used on the distillation columns with the light key and heavy
key impurity mol fractions in respectively the bottoms and distillate as the 2 column specifications. For robust
recycle-tear convergence, the total benzene flow (recycle + fresh) is specified so that the fresh benzene gets
calculated at the end of each recycle tear iteration.

Figure 17.1. Isomerization process schematic with salient design and base operating conditions
17.2. Economic Plantwide Control System (CS1) Design

17.2.1. Step 0: Active Constraint Regions and Economic Operation

The process has 14 independent control valves. Of these, 4 valves must be used to control surge levels,
namely, two reflux drum levels and two sump levels on the columns. Also, two valves will get used to maintain
the columns at their design pressures There are then 8 steady state operating dofs for the process; 1 for the fresh
feed, 2 each for the two columns, 1 for the reactor feed heater, 1 for the reactor effluent cooler and 1 for the
reactor pressure. For robust flowsheet convergence, the chosen 8 specification variables are: the fresh C4 feed
(FC4), the DIB
D1 B1
distillate n-C4 and bottoms i-C4 mol fractions (xnC4 and xiC4 ), the purge column distillate i-C5 D2 B2
and bottoms n-C4 mol fractions (xiC5 and xnC4 ), the reactor inlet temperature (Trxr) and pressure (Prxr) and the
cooler outlet temperature (Tcool).
Of the 8 steady state dofs, Trxr and Prxr are assumed fixed at their design values and not considered for
17
optimization. This is done as the kinetic parameters were adapted by Luyben et al. to match the operating
conditions of an existing industrial reactor and are therefore artificial. Also, in industrial processes, gas phase
reactors are usually operated at the design pressure and not lower so the reaction kinetics are as fast as possible.
Also there is usually a very limited recommended catalyst temperature range for which the technology licensor
guarantees catalyst life. Holding reactor temperature and pressure constant is therefore a reasonable
assumption. The remaining 6 dofs can and should be adjusted for optimizing an economic criterion such as the
steady hourly profit or steam consumption per kg product etc. We consider two process operation modes; Mode
I where the throughput is fixed (eg by market demand-supply considerations) and Mode II where the market
conditions are such that it is optimal to operate the process at maximum throughput.

For Mode I, the optimized economic criterion is the yearly profit, P, defined as
P = [Product Sale – Raw Material Cost – Energy Cost] per year. The fresh C4 feed
(FC4) is fixed at its base case design value (263.1 kmol/h) and the remaining 5 dofs are to be optimized. For
Mode II, the objective is to maximize FC4 using all 6 dofs (including FC4) as decision variables. The
optimization is performed subject to process constraints on the maximum and minimum material / energy
flows, maximum column boilup, maximum product impurity and the maximum allowed reactor temperature.

To simplify the optimization, engineering common sense is applied to reduce the number of decision
B2
variables. To minimize the loss of precious n-C4 down the purge column bottoms, xnC4 is chosen to be small at
1% (base-case design value). In addition, the maximum product
D1 MAX
impurity constraint (xnC4 ) should be active for no product give-away. Finally, the cooler outlet temperature,
Tcool, has almost no impact on the economic objective function and is therefore fixed at a reasonable value of 53
ºC to ensure the reactor effluent vapor is fully condensed using cooling water. These simple engineering
arguments leave 2 decision variables,
B1 D2
xiC4 and xiC5 , for Mode I (FC4 given) optimization. In Mode II (maximum FC4), FC4 is an additional third
decision variable.
The optimization is performed using the fmincon subroutine in Matlab with AspenHysys 2006 as the
background steady state flowsheet solver. The optimization problem formulation and its results are summarized
B2
in Table 17.1. In Mode I, the specified FC4, xnC4 , Tcool , Trxr and Prxr
D1 MAX
values along with xnC4 active constraint leaves two unconstrained steady state dofs B1 D2
corresponding to xiC4 = 0.0565 and xiC5 = 0.02. To maximize throughput (Mode II), these two
unconstrained dofs along with the additional dof corresponding to FC4 are exhausted to drive the
MAXMAX MAX
maximum preheater duty (Qhtr ), maximum purge column boilup (V2 ) and maximum DIB boilup (V1 )
constraints active. At maximum throughput, all steady state dofs get exhausted.

Table 17.1. Isomerization process

Ea i optimization summary
Concentration
i Reaction ki
(kJ/kmol) terms fi(Cj)
1 C6H6 + C3H6 -C9H12 2.8 x 107 104174 CC3CC6
2 C9H12 + C3H6 -C12H18 2.32 x 109 146742 CC3CC9
Objective Maximize(J) J: hourly operating profit *
0 ≤ Material Flows ≤ 2 (base case) 0 ≤ V1, V2, V3 ≤ 1.5 (base case) Vent
Process Constraints Temperature = 32 oC 0 ≤ Energy Flows ≤ 1.7 (base case) 1 bar ≤ PRxr ≤
25 bar Cumene Product Purity ≥ 0.999 mol fraction
Decision Variable Mode I Mode II
FC3 101.93 kmol/h Fixed 169.96 kmol/h
FC6 Total 294.16 kmol/h 316.2 kmol/h
Trxr 322.26 °C 318.58 °C
TRxrShell 368.95 °C 367.98 °C
PRxr 25 bar Max 25 bar Max
Tcooler 100 °C Fixed 100 °C Fixed
Tvent D1 32 °C Fixed 32 °C Fixed
xC3 B1 0.1 % Fixed 0.1 % Fixed
xC9 D2 0.4 % Fixed 0.4 % Fixed
xC6 B2 0.09 % 0.05 %
xC9 D3 99.9 % Min 99.9 % Min
xC9 B3 0.4 % 10 %
Optimum J FC9 Active $3.809x103 h -1 93.59 kmol/h $5.879x103 h -1 150.045 kmol/h xC9 D3 MIN ,
Constraints xC9 D3 MIN , PRxr MAX , V2 MAX PRxr MAX , V2 MAX, V3 MAX
SNo CV Remarks on regulatory / economic significance
Determines
-1-1-1-1-1 process throughput.
*: Heater duty $9.83 GJ; Steam $4.83 GJ;Cooling water $0.16 GJ;FC4 $ 32.5kmol; FiC4 $ 42.0kmol; FiC5 $
1 FC3 -1 Maximum throughput limited by V3 MAX
2 F22.0kmol
C6 Total
&:Increasing
FC4 is specified ; ^: Active constraint; #: FC4 is optimized for maximum throughput
FC6Tot improves selectivity. Maximum FC6Tot limited by V2 MAX .
Affects reactor conversion and selectivity

As the throughput is increased from Mode I (FC4 = 263.1 kmol/h), the optimization of the two
MAX
unconstrained dofs using fmincon shows that Qhtr is the first constraint that becomes active at an FC4 of about
320 kmol/h. A further increase in throughput to 334 kmol/h FC4 drives
MAX MAX
V1 active followed by V2 becoming active at the maximum throughput of 334.5 kmol/h. The increase in
MAX
throughput over what is achieved when Qhtr becomes active is quite small at ~4.5%. We therefore assume
MAX
that once Qhtr becomes optimally active, incrementally higher
MAX MAX
throughput is achieved by driving V1 and V2 constraints active. The large throughput range from 263.1 kmol/h
to the maximum throughput of 334.5
MAX MAX MAX
kmol/h witnesses Qhtr , V1 and V2 becoming active. These constraints are in addition to D1 MAX B2
the other always active constraint xnC4 and specifications for Trxr, Prxr, Tcool and xnC4 , the latter specification
being economically significant. If we assume that Qhtr is adjusted for a desired reactor inlet temperature of Trxr,
MAX
then once the Qhtr constraint becomes active at a high throughput, a further increase in throughput is made
possible by reducing the i-C5 leaking up the purge column distillate and the i-C4 leaking down the DIB column
distillate. The reduced i-C4/C5
circulating around the plant causes the flow through the reactor to reduce allowing more FC4 to be processed
MAX
while keeping the Qhtr constraint active.

17.2.2. Step 1: Loops for Tight Economic CV Control

MAX MAX MAX

At maximum throughput, Qhtr , V1 and V2 are process inputs (potential MVs) constrained to be active.
These are hard active constraints and back-off in these must be
D1 MAX
minimized for process operation at the maximum possible throughput. In addition, xnC4
constraint, which is a process output (CV), is active along with output specifications for Trxr, Prxr, B2 D1MAX B2
xnC4 and Tcool. Of these, tight control of xnC4 and xnC4 is desirable for respectively, on-aim product quality and
B2
small loss of precious n-C4 in the purge column bottoms. The analytical measurement xnC4 is not related to the
product quality and therefore unlikely to be available in practice. As the purge column temperature profile is
quite sharp, the average temperature of
S thth
sensitive stripping tray temperatures, Tpur (14-16tray from top) is therefore controlled as an inferential measure
B2 D1
of xnC4 . Due to their economic significance, we first pair loops for tight control of Qhtr, V2 , V1 and xnC4 at
S
their maximum limits as well as tight control of T pur. MAX
The Qhtr valve is left fully open for process operation at Qhtr . For operating the columns close to
their maximum boil-up limits (i.e. close to flooding limit) with negligible back-off, the respective reboiler
steam valves are used to control the boilups. Thus V1 is paired with Qreb1 and V2 is paired with Qreb2. Tight
D1
control of the product impurity xnC4 is achieved by
MAX S
manipulating the DIB column reflux (L1). Because V2 is active, Tpur cannot be controlled conventionally using
boilup, V2, and the feed to the purge column (B1) is used as the MV instead.
For effective stabilization of the reactor, its pressure and temperature must be controlled tightly. Since
MAX
Qhtr is active, the reactor inlet temperature is maintained at its setpoint using the reactor feed flow stream
(D2). Note that the degree-of-tightness of control in this arrangement would be comparable to Qhtr as the MV
since the open loop dynamic response time constants are likely to be comparable. The reactor pressure is
controlled at its design value by manipulating the cooler outlet valve. To ensure proper condensation of the
reactor effluent, the cooler duty (Qcool) is manipulated to maintain Tcool.

17.2.3. Step 2: Inventory Control System

We now pair loops for remaining inventories that are not important from the economic standpoint. The two
column pressures (Pcol1 and Pcol2) are controlled at their specified values conventionally using the respective
condenser duties (Qcnd1 and Qcnd2). Lastly, we pair loops for the four surge levels on the two columns. Since the purge
column distillate is already paired with the Trxr controller, its reflux drum level (LVLRD2) is controlled using the reflux
rate (L2). The purge column sump level (LVLBot2) is controlled using the column bottoms (B2). Note that even as B2 is
S

a very small stream, effective level control will be achieved as long as T pur is controlled, an economic loop already
paired. The DIB reflux drum level (LVLRD1) is controlled using the distillate (D1). Since B1 is already paired for purge
column temperature control, the DIB column sump level (LVLBot1) is controlled using the fresh C4 feed (FC4). It is
highlighted that in the control structure for maximum throughput operation, the light key i-C4 impurity leaking down
the DIB bottoms and the heavy key i-C5 impurity leaking up the purge column distillate are not

MAX MAX
controlled and float at appropriate values determined by the values of V1 and V2 as well as the other setpoints.
17.2.4. Step 3: Throughput Manipulation and Additional Economic Loops

We now seek an appropriate strategy for reducing throughput while ensuring (near) optimal operation
at lower throughputs. From the optimal Mode I and Mode II results in the
MAX MAX
previous section, V2 is the last constraint to go active. On reducing throughput, V1 is the next constraint to go
MAX
inactive followed by Qhtr . The sensitivity of throughput with respect to the constraint variables decreases in
MAX
order Qhtr, V1 and V2. As explained previously, once Qhtr goes active, only an incremental increase in
MAX
throughput is achieved by reducing the i-C4 leaking down DIB column (this causes V1 to go active) and the
MAX
i-C5 leaking up the purge column (this causes V2 to go active).

The simplest way to reduce throughput (Option 1) would be to maintain the boilups at
MAX MAX MAX
V1 and V2 and reduce Qhtr . Even as throughput would reduce, the operation would be suboptimal due to
overrefluxing in the two columns (unnecessarily high boilups). For near optimal operation at low throughputs,
this overrefluxing must be mitigated. One simple possibility (Option 2) is to hold V2 and V1 in ratio with the
respective column feeds, with the Mode I optimum ratio as their setpoint. Another possibility (Option 3) is to
hold the difference between two appropriate DIB column stripping tray temperatures (ΔTDIB = T37 -T32) constant
by adjusting V1 and holding V2 in ratio with B1. The setpoint for these two controllers would be the Mode I
optimum value. Note that ΔTDIB is controlled instead of a tray temperature as the DIB

B1 D2
temperature profile is quite flat. The last option (Option 4) is to maintain xiC4 and xiC5 at their Mode I optimum
values by adjusting respectively V1 and V2. This however requires two additional composition analyzers, an
unlikely scenario in an industrial setting.
Figure 17.2 compares the optimum steady state profit at various throughputs with the
MAX MAX
profit achieved using the four different options: (1) process operation at V1 and V2 at all MAX MAX
throughputs; (2) V1/(FC4 + D2) and V2/B1 held constant at Mode I optimum till V1 and V2 MAX MAX
become active; (3) ΔTDIB and V2/B1 held constant at Mode I optimum till V1 and V2 B1 D2 MAX MAX
become active and (4) xiC4 and xiC5 held constant at its Mode I optimum till V1 and V2 become active. Note that
for the price data used, the operating profit decreases for a throughput increase beyond FC4 ~332 kmol/h. This
point then represents an economic bottleneck and one would operate below this throughput. The economic
scenario may however change with significantly higher margins for the product, in which case it may become
optimal to operate the process at maximum throughput.

Of the various options considered, Option 4 is economically the best with almost no economic loss
MAX
from optimum till a throughout of FC4 ~ 320 kmol/h, where V1 becomes active. The simpler Option 3 with no
additional composition analyzers is comparable to Option 4. The still simpler Option 3 using ratio controllers
gives slightly higher profit loss (~1%) at low throughputs. The simplest Option 1 is economically the worst
with a significantly higher economic loss between of up to 8% over the throughput range. These results suggest
that Option 2 represents a good compromise between simplicity and minimizing the steady state economic loss.
It is therefore considered for implementation.

The overall throughput manipulation scheme in Option 2 is then as follows. At low throughputs, Qhtr is
MAX
used as the throughput manipulator (TPM). Once Qhtr goes active to increase throughput, throughput
SP
manipulation is shifted to ΔTDIB , which must be increased for a
MAX SP
higher throughput. Once V1 goes active, the TPM is shifted to V2 /B1 , which must again be increased to enhance
MAX
throughput. Once the V2 limit is reached, the process operates at the maximum achievable throughput. A
reverse logic applies for reducing throughput below
 maximum. The TPM for the entire throughput range is then a split range controller, its output
SP SP
shifting from Qhtr to ΔTDIB to V2/B1 to increase throughput from low to maximum and vice

Figure 17.2. Profit for alternative ways of managing the two unconstrained dofs :
Optimum profit ––
B1 D2
: Constant xiC4 and xiC5
––
• • • • : Constant V1/(FC4 + D2) and V2/B1 MAX MAX
• : Process operation at V1 and V2
– –

versa. Figure 17.3 depicts the economic plantwide control structure, labeled CS1 for convenient reference,
including the split-range throughput manipulation scheme. Note that low and high
SP SP
limits are applied on ΔTDIB and V2/B1 for throughput manipulation. The low limit for both SP SP
corresponds to the Mode I optimum values. The high limits for ΔTDIB and V2/B1 are chosen MAX MAX
slightly above the values for which V1 and V2 go active, respectively. In Table 17.2(a), the sequence in which
the different pairings are implemented to obtain CS1 is also listed.

17.3. Conventional Plantwide Control Structure (CS2)

Conventionally, the feed to a process is used as the throughput manipulator and the plantwide control
system is configured with the inventory control loops oriented in the direction of process flow. Such a TPM
choice is often dictated in integrated chemical complexes with the plant feed being set by an upstream process.
Figure 17.4 shows such a conventional plantwide control structure, labeled CS2, for the isomerization process.
To contrast with CS1, the sequence in which the pairings are obtained for CS2 are noted in Table 17.2(b).

In CS2, the column level and pressure controllers are first implemented along with the reactor pressure
and temperature loops (material and energy balance control). On the two
columns, the top and bottom levels are controlled using respectively the reflux and bottoms. The two
column pressures are controlled using the respective condenser duties. The reactor inlet temperature is
controlled using the furnace duty. The reactor pressure is controlled using the reactor effluent condenser
outlet valve while the condensed reactor effluent temperature is controlled using its condenser duty.

Figure 17.3. Economic Plantwide control structure CS1 with split range throughput manipulator for maximum
throughput operation
 Table 17.2. Loop pairing sequence followed for CS1 and CS2

Ea i Concentration
i Reaction ki
(kJ/kmol) terms fi(Cj)
1 C6H6 + C3H6 -C9H12 2.8 x 107 104174 CC3CC6
2 C9H12 + C3H6 -C12H18 2.32 x 109 146742 CC3CC9
Objective Maximize(J) J: hourly operating profit *
0 ≤ Material Flows ≤ 2 (base case) 0 ≤ V1, V2, V3 ≤ 1.5 (base case) Vent
Process Constraints Temperature = 32 oC 0 ≤ Energy Flows ≤ 1.7 (base case) 1 bar ≤ PRxr ≤
25 bar Cumene Product Purity ≥ 0.999 mol fraction
Decision Variable Mode I Mode II
FC3 101.93 kmol/h Fixed 169.96 kmol/h
FC6 Total 294.16 kmol/h 316.2 kmol/h
Trxr 322.26 °C 318.58 °C
TRxrShell 368.95 °C 367.98 °C
PRxr 25 bar Max 25 bar Max
Tcooler 100 °C Fixed 100 °C Fixed
Tvent D1 32 °C Fixed 32 °C Fixed
xC3 B1 0.1 % Fixed 0.1 % Fixed
xC9 D2 0.4 % Fixed 0.4 % Fixed
xC6 B2 0.09 % 0.05 %
xC9 D3 99.9 % Min 99.9 % Min
xC9 B3 0.4 % 10 %
Optimum J FC9 Active $3.809x103 h -1 93.59 kmol/h $5.879x103 h -1 150.045 kmol/h xC9 D3 MIN ,
Constraints xC9 D3 MIN , PRxr MAX , V2 MAX PRxr MAX , V2 MAX, V3 MAX
SNo CV Remarks on regulatory / economic significance
Determines process throughput.
1 FC3 Maximum throughput limited by V3 MAX
2 FC6 Total Increasing FC6Tot improves selectivity. Maximum FC6Tot limited by V2 MAX .
With the basic material/energy
Affects balance loops in place,
reactor conversion pairings for component inventory control are
and selectivity.
implemented3 next. The product
Trxr n-C impurity leaking up the DIB columnFEHE.
Stabilizes reaction heat recycle through
4 is controlled
Affectsby adjusting
reactor D1/L1. The
conversion
ΔTDIB
boilup, V1, is adjusted to maintainand . The purge
selectivity column distillate is maintained in ratio with its reflux while
S S
the bottoms is used control T pur. With the T pur loop, the small purge column bottoms stream would provide
acceptable sump level control. With these pairings, the control structure would provide stable unconstrained
operation ie Mode I operation. The operation would be near optimal for appropriate choice of the steady state
MAX
dof setpoints. Upon hitting constraints such as Qhtr on increasing throughput, appropriate overrides are
needed to ensure control of crucial CVs is not lost. These overrides are also shown in Figure 17.4 and are
briefly explained below.

SP MAX
On increasing the FC4 to increase throughput, the Qhtr constraint would be hit implying loss in control of
Trxr. Losing Trxr control is not acceptable and an alternative manipulation handle for maintaining Trxr is needed.
The closest manipulation handle that would provide tight Trxr control is D2. An override Trxr controller is
therefore implemented with its setpoint slightly below the nominal setpoint. When Qhtr is unconstrained, Trxr
would be above the override controller temperature setpoint and the override controller output would increase.
This output would then be high and the low select block would pass the D2/L2 ratio controller output

SP SP MAX
to D2 (i.e. D2 under ratio control). When Qhtr is hit, Trxr would start decreasing and go below the override
controller setpoint,Figure
whose17.4.
output would decrease
Conventional till thecontrol
plantwide low select ultimately
structure, CS2
SP SP
passes this signal to D2 (i.e. D2 under Trxr control).
(a) Basic pairings for Mode I (unconstrained) operation
(b) Overrides for handling constraints
 MAX
It is possible to bring about a near optimal increase in throughput with Qhr active by driving
MAX MAX V1 and V2 active, in that order. To do so, a PI Qhtr override controller with its setpoint very close to
MAX
the Qhtr limit is implemented. The high select on the Qhtr override output and the ΔTDIB controller output,
selects the greater of the two signals. The selected signal is sent as the
SP MAX
setpoint to the V1 controller through a low select that ensures V1 does not ever exceed V1 . At low throughputs
MAX
(FC4 low, Qhtr < Qhtr ) the direct acting Qhtr override controller output would decrease and the high select
would pass the ΔTDIB controller output. On sufficiently increasing FC4, Qhtr would increase above the override
SP
controller setpoint, and the controller output would start to increase. The high select would ultimately pass V1
manipulation to the Qhtr override,
SP SP
which would cause V1 to increase. If FC4 is high enough (or increased fast enough), V1 would MAXS SP
reach V1 . The Tpur controller would increase V2 to ensure that the n-C4 does not leak out the purge column
MAX
bottoms. V2 going active would signal that fresh n-C4 beyond the processing capacity of the plant is being fed.
To automatically reduce FC4 to the maximum processing capacity limit, an override scheme for altering the
MAX
material balance structure from V2 all the way back to the process fresh feed is implemented.

MAX S
When V2 goes active, Tpur control is lost implying excessive leakage of precious n-C4 down the purge
column bottoms and consequent economic loss. To prevent the same, an alternative manipulation handle for
S

T pur is needed. The feed to the purge column would provide reasonably tight tray temperature control. A PI
S MAX

T pur override controller with its setpoint slightly below the nominal setpoint is implemented. When V2 is
inactive, the tray temperature would be higher than the override controller setpoint so that its output would
S

increase. The low select on LVLBot1 controller output and the T pur override controller output would pass the
former signal
SP MAX S
to B1 (purge column feed under LVLBot1 control). When V2 goes active, Tpur control would be lost and it would
decrease below the override controller setpoint. The controller output would then decrease and the low select
SP S

would ultimately pass B1 manipulation to the override controller (purge column feed under T pur control).
LVLBot1 control is now lost and it would increase. A reverse acting LVLBot1 override controller with a setpoint
slightly higher than the nominal setpoint is implemented. As LVLBot1 increases, its output would decrease
SP

(reverse action). The low select on this signal and operator specified FC4 would ultimately pass the former
signal as the setpoint to the fresh C4 feed flow controller causing the fresh feed to be cut

MAX 21
by the appropriate amount once V2 goes active. As recommended by Shinskey , external reset on all PI
controllers whose output passes through a low/high select block is used to ensure that when inactive, the output
is not too far from the selected signal due to reset windup. This ensures quick 'taking over' of control so that the
duration for which a CV remains unregulated is as small as possible. The external reset is implemented
internally in AspenHysys.

17.4. Dynamic Simulations and Closed Loop Results

17.4.1. Tuning of Controllers

The performance of the two control structures, CS1 and CS2, is evaluated using rigorous dynamic
simulations in AspenHysys 2006. To ensure that any differences in the performances are largely attributable to
the structure, a consistent tuning procedure is followed for tuning the loops in both the structures. All flow
controllers are tuned with a gain of 0.5 and a reset time of
0.5 mins. All pressure controllers are tuned for tight pressure control, which is any way necessary for
stabilizing the pressure driven dynamic simulation. All level controllers are P only
 with a gain of 2. The only exception is the DIB sump level controller in CS1 where a lower gain of 1 is used
since the lag between the sump and the fresh C4 feed is significant due to the intervening 20 stripping trays. In
all temperature loops, the temperature measurement is lagged by 1 minute to account for sensor dynamics.
Also, the controller output signal is lagged by 2 mins to account for heat transfer equipment dynamics. The only
exception is the cooler temperature controller where a higher 8 min lag is applied to account for the slow
dynamics of a flooded condenser. All temperature controllers are PI(D) and tuned using the autotuner with
minor refinement for a not-too-oscillatory closed loop servo response, if necessary. In the PI product
composition control loop, a 5 minute dead time and a 5 minute measurement sampling time is applied. The
autotuner does not give reasonable tuning and the open loop step response is used to set the reset time at the
rd
2/3 response completion time and the controller gain adjusted for a not-too-oscillatory servo response. In both
structures, the product composition loop is tuned first with the ΔTDIB loop on manual followed by tuning of the
ΔTDIB loop with the composition loop on automatic. This ensures that all the detuning due to multivariable
interaction gets taken in the ΔTDIB loop and not the product purity loop. This gives tight product purity control,
an economically important control objective.

S
In the CS2 override scheme, the override setpoint for Trxr and T pur cannot be chosen too close to the
corresponding nominal controller setpoint as that would lead to unnecessary controller output switching during
routine transients causing further transients. Accordingly the override controller setpoint is chosen as close as
possible to the corresponding nominal controller setpoint for the disturbance that causes the worst-case
transients. It is also highlighted that the Qhtr override controller that manipulates V1 is a long loop with slow
MAX
dynamics. Since its setpoint must be close to Qhtr , a P only controller would require a large gain to ensure V1
MAX
gets driven to V1 for achieving maximum throughput. The large gain leads to on-off control for routine
disturbances at a high but below maximum throughput with the override taking over and giving up V1
manipulation. A loose PI Qhtr override controller is therefore implemented to ensure its setpoint is close to Qhtr
MAX
and on-off control is avoided. Table 17.3 lists the salient controller tuning parameters for CS1 and CS2 using
the above procedure.

Table 17.3. Controller tuning parameters

Regulatory
Ea i controllers
Concentration
i Reaction ki
(kJ/kmol) terms fi(Cj)
1 C6H6 + C3H6 -C9H12 2.8 x 107 104174 CC3CC6
2 C9H12 + C3H6 -C12H18 2.32 x 109 146742 CC3CC9
Objective Maximize(J) J: hourly operating profit *
0 ≤ Material Flows ≤ 2 (base case) 0 ≤ V1, V2, V3 ≤ 1.5 (base case) Vent
Process Constraints Temperature = 32 oC 0 ≤ Energy Flows ≤ 1.7 (base case) 1 bar ≤ PRxr ≤
25 bar Cumene Product Purity ≥ 0.999 mol fraction
Decision Variable Mode I Mode II
CS2 override controllers
MV KC τi (mins) SP Sensor Span
Qhtr V1 0.05 150 1230kW
o
0-1294kW
o
Trxr
D1 0.4 10 199.5
o
C 160-240
o
C STpur
B1 0.5 40 58 C 40-80 C LVLBot1
FC4 2 -70% 0-50%
17.4.2. Closed Loop Results

The plantwide transient response of the two control structures, CS1 and CS2, is obtained for principal
disturbances for Mode I and Mode II operation. In Mode I, a ±5 mol% step change in the fresh C4 feed i-C4 mol
fraction with a complementary change in the n-C4 mol fraction and a ±20 kmol/h FC4 (throughput change) are
considered the principal disturbances. In Mode II, only the feed composition step change is considered the
principal disturbances as the throughput gets fixed by the active constraints. The dynamic response is also
obtained for a throughput transition from Mode I to Mode II and back.

Figure 17.5 plots the dynamic response of salient process variables to a feed composition step
disturbance in Mode I for CS1 and CS2. Both structures are observed to effectively reject the disturbance with
tight control of the n-C4 impurity in the product. In CS1, FC4 gets adjusted and the flow to the reactor settles to
the appropriate value for maintaining Trxr for the set Qhtr, the TPM. In CS2 on the other hand, the FC4 (TPM)
remains fixed and the i-C4 production changes in proportion to the n-C4 in the fresh feed. In both structures, the
S
leakage of n-C4 down the purge column bottoms is well regulated via the action of the T pur controller.

Figure 17.5. Mode I transient response to ±5% feed n-C4 composition change : +5% CS1
––
: -5% ––––
––
CS1 : +5% CS2 : -5% CS2
 Figure 17.6 plots the Mode I dynamic response for a ±20 kmol/h change in FC4. In CS1, the Qhtr
setpoint must be increased (decreased) by 169 kW (~21% of base-case Qhtr) to bring about a 20 kmol/h (~7.6%
of base-case FC4) increase in FC4. Similarly, Qhtr must be decreased by 138 kW (~17.2%) for achieving the
decrease in FC4. For the throughput change disturbance, the product impurity is well controlled in the transient
period in both CS1 and CS2. The transient deviations in CS1 are slightly lower than in CS2 due to more severe
transients in the recycle loop in the latter. In CS1 on the other hand, the recycle loop transients are less severe
(smooth response). In addition to tight product impurity control, both the structures achieve tight regulation of
S
the n-C4 leakage in the purge column bottoms via the action of the T pur controller. The transient variability in
B2
xnC4 is significantly higher in CS1 as a large change in Qhtr (TPM) causes a large change in D2 which severely
disturbs the purge column material balance.

Figure 17.6. Mode I transient response to ±5% feed n-C4 composition change –– : +5%
CS1 –– : -5% CS1 –– : +5% CS2 –– : -5% CS2
 Figure 17.7 plots the Mode II dynamic response to a ±5% feed composition step change. All override
controllers in CS2 are active so that structurally, CS1 and CS2 are very similar. The only significant difference
S
is that the setpoint of the Trxr and T pur override controllers in CS2 is slightly lower than the corresponding
nominal setpoint values. In CS1, on the other hand, the setpoint values are held at their nominal values. As seen
from the dynamic responses, the plantwide transient response is smooth in both the structures. Also, tight
control of product impurity and the n-C4 leakage down the purge column bottoms is achieved. In CS2 however,
the production of i-C4 at the initial and final steady state is slightly lower than in CS1 due to the slightly lower
Trxr setpoint which causes a slight reduction in single pass reactor conversion as well as higher n-C4 leakage in
S
the purge column bottoms due to the lower T pur setpoint.

–– ––––
The synthesized control structures are also tested for a large throughput transition from the design throughput
(FC4 = 263.1 kmol/h) to the maximum achievable throughput and back. The transient response is shown in
Figure 17.8. In CS1, to increase throughput, the split range
SP SP SP
scheme switches the TPM from Qhtr to ΔTDIB and then to V2/B1 . The switching order gets reversed for
decreasing the throughput. The transient response shows that tight product impurity
control is achieved across the entire throughput range. The loss of precious n-C4 down B2 is also regulated at a
small value. Most importantly, the plantwide transients are smooth and not too severe.

Figure 17.8. Throughput transition for CS1 and CS2 : CS1 :

steady throughput MAX MAX

CS2
–– –– MAX MAX

SP
In CS2, FC4 is ramped up causing Qhtr to increase and as it crosses the Qhtr override controller setpoint
MAX SP
(chosen setpoint is 95% of Qhtr ), the override output increases above the ΔTDIB controller output passing V1
manipulation to the Qhtr override, which slowly keeps on
SP MAX MAX
increasing V1 to V1 . As and when Qhtr is reached, Trxr decreases and the override Trxr SP MAX
controller takes over D2 manipulation. Meanwhile, V2 increases rapidly and hits V2 as more n-C4 is being fed in
S
than being consumed in the reactor. This causes T pur to decrease and the override scheme for altering the
material balance structure gets activated to cut the FC4 feed. Since the Qhtr override is a long loop, the increase
MAX
in V1 is slow and even after 75 hrs, the V1 constraint is not approached and the product rate, D1, is about 299
kmol/h (~20 kmol/h < maximum steady D1). Even as D1 reaches its maximum steady value, it takes a very long
SP
time. After 75 hrs, FC4 is ramped down to its base value (263.1 kmol/h) and a smooth transition
 MAX S
occurs. From the CS2 response in Figure 17.8, notice that in the small period where V2 goes active and T pur
override starts manipulating B1, large transient loss of precious n-C4 down the purge column bottoms occurs.
S
Also, the steady n-C4 loss is higher due to the lower than nominal setpoint of the T pur override.

Another pertinent comparison is the transients caused due to overrides taking over / giving up control
during routine disturbances. We consider a worst-case step disturbance in the fresh feed composition, where the
n-C4 composition increases by 5% with initial steady operation at FC4 = 293.1 kmol/h, where none of the
constraints are active. The transient response of CS1 and CS2 to this disturbance is shown in Figure 17.9. CS1
effectively rejects the disturbance with tight product purity control and regulation of n-C4 in the purge column
bottoms with the plant settling down at the new steady state in about 30 hrs. In CS2, on n-C4 composition
12
increasing by 5%, a large transient increase occurs in Qhtr due to the snowball effect , which triggers the Qhtr
MAX
override. V1 is then slowly driven towards V1 while the additional n-C4 causes V2 to increase. The slow
increase in V2 causes the i-C5 circulating in the plant and hence D2 to decrease. For the lower D2 (reactor feed),
Trxr control eventually passes back to Qhtr and the plant settles at the new steady state in about 75 hrs, which is
more than twice the time for CS1. If the Qhtr override controller is made aggressive by increasing the
proportional gain by a factor of 2, oscillations due to the Trxr override successively going active and inactive are
observed (see Figure 17.9). The dynamic performance thus degrades significantly at high throughputs where
the overrides get activated. It is then not surprising at all that operators tend to switch the overrides off and
make the necessary adjustments manually.

17.4.3. Quantitative Economic Performance Comparison

A quantitative economic comparison of the two control structures is performed for maximum
throughput (Mode II) operation. We consider a +5% feed n-C4 composition step change as the worst case
disturbance. Table 17.4 compares the maximum achieved steady throughput (FC4) along with the corresponding
n-C4 component flow (loss) in the purge column bottoms, the i-C4 product rate and the operating yearly profit
for CS1 and CS2. Expectedly, no back-off and throughput loss is observed for CS1, which has been designed
for process operation with all the hard active constraints at their maximum limits. In contrast, in CS2, due to the
S 6
need for the Trxr and T pur override setpoints to be lower than nominal, an yearly profit loss of $0.45x10 (~2%)
occurs compared to CS1. The override controller setpoint offsets have been chosen to be as small as possible at
S
1 °C for Trxr and 5 °C for T pur to ensure that the overrides do not get triggered during routine transients. CS2,
which was obtained without any consideration of the constraints that go active at higher throughputs, thus is
economically and dynamically inferior to CS1 regardless of the approach used to handle constraints (back-off
or overrides). Overall, these results demonstrate that the full active constraint set plays a key role in economic
plantwide control system design.

17.5. Conclusions

In conclusion, this case study on plantwide control of the C4 isomerization process demonstrates that a simple
decentralized plantwide control system for achieving near optimal and smooth process operation over a wide
throughput range can be synthesized. The active constraints at maximum throughput form the key to devising
the control system. These
constraints dictate the pairings for tight control of these active constraints and the consequent pairings for
inventory regulation as well as the throughput manipulation strategy. Quantitative results show that a
conventional control structure with the TPM at the process feed with overrides for handling constraints is
economically inferior with a steady profit loss of ~2% at maximum throughput due to the offset needed in
the override controller setpoints. The conventional scheme is also found to be dynamically inferior. The
case study demonstrates the crucial role of the active constraints in economic plantwide control structure
synthesis.

310 1300

T
C nC
4
1
F Vk2
kr
rx mht
mC ol/r
ol4 hr2
/h Q9
r 0 20 40 60 80 100 k0 0 20 40 60 80 100 300 1200
W1
1
2000 0 1000
0
2
o
8
V 0
k 1
m 0
ol/ 0
hr 0
2
0 20 40 60 80 100 7 0 20 40 60 80 100 1900 1800
0 1700 800 600 400 1600 200
9
201 0 0.022
0
0.021
200
D
1x

0 20 40 60 80 100 Time hr 0 20 40 60 80 100 0.02 199

0.019 198 0.018 Time hr

Figure 17.9. Transient response for +5% feed n-C4 composition change at FC4 = 293.1kmol/hr : CS1 : CS2
–– –– ––
aggressive Qhtr override : CS2 loose Qhtr override

Table 17.4. Mode II throughput loss comparison for +5 mol% feed composition step change
CS1 CS2
Ea i Concentration
i Reaction ki
(kJ/kmol) terms fi(Cj)
1 C6H6 + C3H6 -C9H12 2.8 x 107 104174 CC3CC6
2 C9H12 + C3H6 -C12H18 2.32 x 109 146742 CC3CC9
Objective Maximize(J) J: hourly operating profit *
0 ≤ Material Flows ≤ 2 (base case) 0 ≤ V1, V2, V3 ≤ 1.5 (base c
Bibliography
1. Al-Arfaj MA, Luyben WL. Plantwide control for TAME production using reactive distillation. AIChE J.
2004;50(7):1462-1473.
2. Araujo A, Skogestad S. Control structure design for the ammonia synthesis process. Comp Chem Eng.
2008;32(12):2920-2932.
3. Aske EMB, Strand S, Skogestad S. Coordinator MPC for maximizing plant throughput. Comp Chem Eng.
2008;32:195-204.
4. Aske EMB, Skogestad S. Consistent inventory control. Ind Eng Chem Res. 2009;48(44): 10892-10902.

5. Douglas JM. Conceptual Design of Chemical Processes. New York: McGraw Hill, 1988.
6. Gera V, Kaistha N, Panahi M, Skogestad S. Plantwide control of a cumene manufacture process. Computer
Aided Chemical Engineering, 2011, 29 (8), 522-526.
7. Jagtap R, Kaistha N, Skogestad S. Plantwide control for economic optimum operation of a recycle process
with side reaction. Ind Eng Chem Res. 2011;50(14):8571-8584.
8. Jagtap R, Kaistha N. Economic plantwide control of the ethyl benzene process. AIChE J. 2012; submitted.

9. Jagtap R, Kaistha N. Economic plantwide control of the C4 isomerization process. Ind Eng Chem Res. 2012;
submitted.
10. Jagtap R, Kaistha N. Throughput manipulator location selection for economic plantwide control. In
Rangaiah GP, Kariwala VA. Advances in Plantwide Control. Upper Saddle River NJ: John Wiley and Sons,
2012 in press.
11. Kanodia R, Kaistha N. Plantwide control for throughput maximization: A case study. Ind Eng Chem Res.
2010;49(1):210-221.
12. Luyben M.L, Luyben W.L. Design and control of a complex process involving two reaction steps, three
distillation columns, and two recycle streams. Ind. Eng. Chem. Res. 1995, 34 (11), 3885-3898.

13. Luyben M.L, Tyreus B.D, Luyben W.L. Analysis of control structures for reaction/separation/recycle
processes with second order reactions. Ind. Eng. Chem. Res. 1996, 35 (3), 758-771.

14. Luyben ML, Tyreus BD, Luyben WL. Plantwide control design procedure. AIChE J.
1997;43(12):3161-3174.
15. Luyben WL. Snowball effects in reactor/separator processes with recycle. Ind Eng Chem Res.
1994;33(2):299-305.
16. Luyben WL, Tyreus BD, Luyben ML. Plantwide Process Control. New York: McGraw Hill, 1999.

17. Luyben WL, Tyreus BD, Luyben ML. Isomerization process, Plantwide Process Control, McGraw Hill:
New York, 1999, 273-293.
18. Luyben WL. Plantwide control of an isopropyl alcohol dehydration process. AIChE J.
2006;52(6):2290-2296.
19. Luyben WL. Design and control of the cumene process. Ind. Eng. Chem. Res. 2010, 49 (2),
719.
20. Price RM, Lyman PR, Georgakis C. Throughput manipulation in plantwide control structures. Ind Eng
Chem Res. 1994;33(5):1197-1207.
21. Shinskey FG. Process Control Systems: Application, Design and Tuning. New York: McGraw Hill, 1996.
22. Singh S, Lal S, Kaistha N. Case study on tubular reactor hot-spot temperature control for throughput
maximization. Ind Eng Chem Res. 2008;47(19):7257-7263.
23. Skogestad S. Plantwide control: The search for the self-optimizing control structure. J Proc Cont. 2000,
10(5), 487-507.
24. Skogestad S. Control structure design for complete chemical plants. Comp Chem Eng.
2004;28(1-2):219-234.
25. Skogestad S. Do's and dont's of distillation control. Chem Eng Res Des (Trans I ChemE, Part A).
2007;85(A1):13-23.
26. Tyreus BD, Luyben WL. Dynamics and control of recycle systems 4. Ternary systems with one or two
recycle stream. Ind. Eng. Chem. Res. 1993, 32 (6), 1154-1162.
27. Zhang C, Vasudevan S, Rangaiah GP. Plantwide control system design and performance evaluation for
ammonia synthesis process. Ind Eng Chem Res. 2010;49(24):12538-12547.

..... .....

..... .....
....................................................................................

.....................................................

.........................................................................................................................
. . . . . ...... ................................................... .....

.....................................-..............................-........-........-........-...

.....................................-............................-........-......-......-...
...............................................................................................................
..................................................................
..................................................................
..................................... -.................................. -....... -
....................................................
䘀椀最甀爀攀..................................................................
..................................................
....................................................................................
 PRACTICE PROBLEMS FOR THE COURSE
Note: Many of these problems have been taken from “Process Modeling, Simulation and Control for
Chemical Engineers” by William L Luyben.

PROBLEMS FOR MODULE I

I.1 For the dynamic responses to a unit step in the input (see Figure), state reasonable transfer
functions that would fit well to the shown response.

4
1
3
2
2

3
1

0
7
-1

-2 5
4
-3
6
-4
0 2 4 6 8 10 12 14 16 18 20

I.2 Consider the series system in Figure.

u yS yM

a) Design a conventional P only feedback controller for a closed loop damping coefficient of 0.5.
b) Design a conventional PI feedback controller for a closed loop damping coefficient of 0.5. Choose
the reset time to be the dominant open loop time constant.
c) Design a series cascade control system. The slave loop is P only and tuned for critical damping. The
master loop is PI and tuned for closed loop damping coefficient of 0.5.

I.3 Consider a process described by the transfer function

 under P control. What value of KC gives LCLMAX = 2dB.

I.4 Consider the log-modulus of a second order-underdamped transfer function of unit gain. Derive
and analytical expression between its maximum log-modulus and damping coefficient (ε). Tabulate
values of maximum log modulus for ε = 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7. What is the damping
coefficient for a maximum log modulus of 2 dB.

I.5 Consider a first order lag plus deadtime process controlled by a P controller. Devise a procedure for
obtaining KC for LCLMAX = 2 dB.

I.6 Find KC for which LCLMAX = 2 dB for a process with transfer function

and controlled by a P controller.

I.7 For the process transfer function in question V, calculate KU and PU and subsequently the ZN
settings for a P, PI and a PID controller. Also tabulate the TL settings for a PI and a PID controller.

I.8 A process has the following open loop transfer function relating controlled (y) and manipulated (u)
variables:

a) Is the open loop system stable. Explain why.

b) If a proportional feedback controller is used, what is the largest ratio of D/τ for which a phase
margin of 45° can be obtained.

I.9 Consider a 3x3 open loop multivariable system

(a) Calculate its RGA. Based on the RGA, what input-output pairing(s) would you recommend.
(b) Tabulate the Niederlinski Index for all the possible input-output pairings. What can you say about
the closed-loop integral stability of the pairing(s) recommended in (a).
(c) For the recommended pairing, design a feedforward dynamic decoupler showing its complete block
diagram and also the realizable feedforward compensator transfer functions.
PROBLEMS FOR MODULE II

II.1 Sketch a distillation column with a vapor sidestream from the stripping section and an intermediate
reboiler that vaporizes liquid from a stripping tray located immediately above the vapor side draw
tray. The vapor from the intermediate reboiler is fed back into the column immediately above the
tray from where the liquid is drawn into the intermediate reboiler. The feed to the column contains
non-condensables which are vented from the reflux drum as vapor distillate. The usual liquid
distillate and bottoms are withdrawn from the column. The intermediate reboiler uses low-
pressure steam while the bottom reboiler uses high-pressure steam. The condenser uses cooling
water for heat removal.
a) Draw all independent control valves for the column described above
b) Excluding the feed and column operating pressure, what is the steady state operating degree of
freedom of this column.
(Hint: Count total number of valves and discount valves used for feed flow, column pressure and
surge level controllers)
c) Draw a control concept diagram in which
• Column top pressure is controlled by adjusting the vapor vent valve.
• Reflux drum level is controlled by distillate flow.
• Reflux is ratioed to feed rate
• Low pressure steam to intermediate reboiler is ratioed to the column feed.
• Vapor side-draw is adjusted to hold a stripping tray temperature constant.
• Bottom sump level is controlled by adjusting high pressure steam flow to the bottom reboiler.
• Bottoms is flow controlled
• Low sump level overrides the low pressure steam flow to the intermediate reboiler.
• High column pressure overrides both steam valves.

II.2 Gas feed to an adiabatic tubular reactor is heated to 700 °F in a gas-fired furnace. The reaction is
endothermic. The exit temperature of the gas leaving the reactor is to be maintained at 600 °F.
a) Sketch the process flow sheet with a valve on the gas feed and another valve on the furnace fuel
gas line.
b) Draw a control concept diagram that accomplishes the following:
• Gas feed is flow controlled
• Fuel gas to the furnace is flow controlled and ratioed to the gas feed rate.
• The fuel to feed ratio is adjusted by a furnace exit temperature controller.
• The setpoint of the furnace exit temperature controller is adjusted by a reactor exit temperature
controller.
• Furnace exit temperature controller setpoint is not to exceed 750 F
• High furnace stack-gas temperature should pinch the fuel gas control valve.

II.3 The Petlyuk column configuration for separating a ternary mixture into its nearly pure constituent
components is shown in the Figure.
a) Draw all independent control valves for the process
b) Excluding the throughput and column pressures, what is the steady state operating degrees of
freedom for the process.
c) Based on your understanding of the function of each of the 6 column sections, develop a
temperature inferential control structure for the Petlyuk column configuration.

A
II.4 A compressor pressurizes gas in a gas-header, which supplies multiple downstream users. When
the downstream users’ demand is low, the compressor drive speed is adjusted to maintain the
surge drum (or header) pressure. If however, the demand exceeds a certain maximum limit, to
avoid overloading the compressor drive, the drive speed is adjusted to maintain the flow through
the compressor at the maximum limit. Draw an override control system that accomplishes the
above control objectives.
FC
User 1
FC
User 2
Header

RPM Surge
FC
Drum
User N-1
FC
User N

Figure. Gas compression unit for downstream users

II.5 Sketch a control concept diagram for the exothermic chemical reactor (see Figure) with reaction
heat being used to generate steam.
• Steam drum pressure is controlled by the steam valve
• Condensate flow is ratioed to feed flow
• Steam drum liquid level is controlled by adjusting the condensate to steam ratio.
• Feed is flow controlled
• Reactor liquid level is controlled product withdrawal.
• Reactor temperature is controlled by adjusting the stem pressure controller setpoint
• A high reactor temperature override pinches the reactor feed
• A low steam drum level override pinches the reactor feed

Steam

Condensate

Feed

Product

II.6 Sketch the following control concept diagrams

• Temperature in a reactor is controlled by manipulating cooling water to a cooling coil. Pressure in
the reactor is controlled by admitting a reactant gas feed into the reactor. A high temperature
override pinches the gas feed valve.
• The reflux drum level is controlled by the reflux flow back into the column. Distillate is maintained
in ratio with the reflux. The ratio setpoint is adjusted by a rectifying tray temperature controller.
The temperature controller setpoint is adjusted by a distillate composition controller.

II.7 Sketch a control concept diagram for the distillation column shown in Figure with the following
information:
• Reflux is flow controlled, ratioed to feed rate, and over-ridden by low reflux drum level.
• Steam is flow controlled, with the flow controller setpoint coming from a temperature controller
that controls a tray temperature in the stripping section of the column. Low base level or high
column pressure pinches the steam valve.
• Base level is controlled by bottom product flow rate.
• Reflux drum level is controlled by distillate product flow rate.
• Column pressure is controlled by changing the setpoint of the speed controller on the compressor
turbine. The speed controller output sets a flow controller on the high pressure steam to the
turbine.
• A minimum flow controlled (“anti-surge”) sets the valve in the compressor bypass line to prevent
the flow-rate through the compressor from dropping below some minimum flowrate.
 Compressor

Bypass
HP Steam

Distillate
Reflux
Feed

Steam
Problems for Module III

III.1 CO2 in stack gases is usually removed via a closed circuit amine absorption-desorption process
shown in the Figure. The CO2 rich stack gas is fed to an absorption tower where a CO2 lean aqueous
amine solution absorbs the CO2. The CO2 lean gas is discharged to the atmosphere. The CO2 rich
amines stream from the absorber is preheated in a process-to-process heat exchanger and fed to a
stripping column that uses steam reboiler to strip out the absorbed CO2. The hot vapors from the
stripper are condensed with the condensed liquid being refluxed back to the column (total reflux)
and the non-condensable CO2 with some water vapor being vented as vapor distillate.
a) For this process, design a basic conventional regulatory control system with the gas feed as the
throughput manipulator (TPM).
b) If the bottleneck constraint is the flooding limit on the stripping column, then the stripper boilup
cannot exceed a maximum limit. On hitting this capacity constraint, all the CO2 absorbed in the
absorber cannot be boiled off in the stripper and an override system is needed that would cut the
gas-feed. Design appropriate overrides on top of the regulatory control system synthesized in part
(a) for accomplishing the same.
c) Use the stripper vapor boilup as the TPM and configure a control system that does not require any
overrides for handling the above bottleneck constraint.

CO2 Out
CO2 Lean Cooler
Gas Out
Absorber

Stripper

PPHE

CO2 Rich
Gas In Steam
CO2 Rich CO2 Lean
Amines Amines

Figure. CO2 sequestration process

III.2 The process flowsheet producing cumene from benzene and propylene is shown in Figure. The
reactions are
MAIN: Benzene + Propylene → Cumene
SIDE: Cumene + Propylene → Diisopropyl Benzene (DIPB)
TRANSALKYLATION: DIPB + Benzene ↔ 2 Cumene

a) Rationally draw all independent valves for the process.

b) For given throughput, reactor pressure and column pressures, what is the steady state operating
degrees of freedom for the process. State natural specifications for these degrees of freedom.

c) Using the independent control valves, propose a plant-wide control structure for the process with
the fresh limiting reactant feed as the throughput manipulator.
d) If the reactor cooling duty maxing out is the bottleneck constraint, devise appropriate overrides for
handling the constraint as throughput is increased.
e) Revised the control structure with the reactor cooling duty as the TPM.

Process Description:

The reaction chemistry is as above. Fresh benzene (pure, liquid), fresh propylene (liquid containing 5%
inert propane) and recycle benzene are mixed. The mixed stream is vaporized in the vaporizer,
preheated in a feed effluent heat exchanger, heated to the reaction temperature in a furnace and then
fed to the catalytic reactor. The highly exothermic main and side reactions occur in the reactor, which is
a cooled shell and tube heat exchanger with catalyst loaded tubes. The reaction heat is removed as
steam. Almost complete conversion of propylene occurs in the reactor. The reactor effluent is
depressurized across the PRV (pressure reducing valve) and cooled in the cooler. The gases (unreacted
propylene and propane) are removed as fuel gas vapor distillate from a purge column. The purge
column bottoms is fed to the recycle column that separates the unreacted benzene from the reaction
products (cumene and DIPB). The distillate stream, which is essentially pure benzene, is recycled while
the bottoms is fed to the product recovery column that separates cumene and DIPB as the distillate and
bottoms respectively. The heavy DIPB is mixed with a small amount of recycle benzene and sent to an
adiabatic transalkylator after preheating. The reversible transalkylation reaction occurs in the
transalkylator to convert the DIPB back to cumene. For high equilibrium conversion of DIPB, the
benzene to DIPB ratio is maintained large. The transalkylator effluent contains cumene, the extra
benzene and the unconverted DIPB. It is fed to the recycle column for separating the three components.
At steady state, the net DIPB formation by the side reaction equals the net DIPB consumption by
transalkylation. In other words, the DIPB is allowed to accumulate in the DIPB recycle loop to an extent
such that the net DIPB generation rate is zero. The DIPB is thus recycled to extinction.
PROBLEMS FOR MODULE IV

IV.1 For the recycle process, ethyl benzene process and cumene process case studies that we have gone
over in Module IV, design a control structure for maximum reactor heat removal as the bottleneck
constraint.

IV.2 For the zero DIPB discharge cumene process in Problem III.2, the zero furnace duty, maximum
reactor hot spot temperature and maximum recycle column boilup are active hard capacity
constraints at maximum throughput. Design an economic plantwide control structure in light of
these capacity constraints.
Bibliography
1. Al-Arfaj MA, Luyben WL. Plantwide control for TAME production using reactive distillation. AIChE J.
2004;50(7):1462-1473.
2. Araujo A, Skogestad S. Control structure design for the ammonia synthesis process. Comp Chem Eng.
2008;32(12):2920-2932.
3. Aske EMB, Strand S, Skogestad S. Coordinator MPC for maximizing plant throughput. Comp Chem Eng.
2008;32:195-204.
4. Aske EMB, Skogestad S. Consistent inventory control. Ind Eng Chem Res. 2009;48(44): 10892-10902.

5. Douglas JM. Conceptual Design of Chemical Processes. New York: McGraw Hill, 1988.
6. Gera V, Kaistha N, Panahi M, Skogestad S. Plantwide control of a cumene manufacture process. Computer
Aided Chemical Engineering, 2011, 29 (8), 522-526.
7. Jagtap R, Kaistha N, Skogestad S. Plantwide control for economic optimum operation of a recycle process
with side reaction. Ind Eng Chem Res. 2011;50(14):8571-8584.
8. Jagtap R, Kaistha N. Economic plantwide control of the ethyl benzene process. AIChE J. 2012; submitted.

9. Jagtap R, Kaistha N. Economic plantwide control of the C4 isomerization process. Ind Eng Chem Res. 2012;
submitted.
10. Jagtap R, Kaistha N. Throughput manipulator location selection for economic plantwide control. In
Rangaiah GP, Kariwala VA. Advances in Plantwide Control. Upper Saddle River NJ: John Wiley and Sons,
2012 in press.
11. Kanodia R, Kaistha N. Plantwide control for throughput maximization: A case study. Ind Eng Chem Res.
2010;49(1):210-221.
12. Luyben M.L, Luyben W.L. Design and control of a complex process involving two reaction steps, three
distillation columns, and two recycle streams. Ind. Eng. Chem. Res. 1995, 34 (11), 3885-3898.

13. Luyben M.L, Tyreus B.D, Luyben W.L. Analysis of control structures for reaction/separation/recycle
processes with second order reactions. Ind. Eng. Chem. Res. 1996, 35 (3), 758-771.

14. Luyben ML, Tyreus BD, Luyben WL. Plantwide control design procedure. AIChE J.
1997;43(12):3161-3174.
15. Luyben WL. Snowball effects in reactor/separator processes with recycle. Ind Eng Chem Res.
1994;33(2):299-305.
16. Luyben WL, Tyreus BD, Luyben ML. Plantwide Process Control. New York: McGraw Hill, 1999.

17. Luyben WL, Tyreus BD, Luyben ML. Isomerization process, Plantwide Process Control, McGraw Hill:
New York, 1999, 273-293.
18. Luyben WL. Plantwide control of an isopropyl alcohol dehydration process. AIChE J.
2006;52(6):2290-2296.
19. Luyben WL. Design and control of the cumene process. Ind. Eng. Chem. Res. 2010, 49 (2),
719.
20. Price RM, Lyman PR, Georgakis C. Throughput manipulation in plantwide control structures. Ind Eng
Chem Res. 1994;33(5):1197-1207.
21. Shinskey FG. Process Control Systems: Application, Design and Tuning. New York: McGraw Hill, 1996.
22. Singh S, Lal S, Kaistha N. Case study on tubular reactor hot-spot temperature control for throughput
maximization. Ind Eng Chem Res. 2008;47(19):7257-7263.
23. Skogestad S. Plantwide control: The search for the self-optimizing control structure. J Proc Cont. 2000,
10(5), 487-507.
24. Skogestad S. Control structure design for complete chemical plants. Comp Chem Eng.
2004;28(1-2):219-234.
25. Skogestad S. Do's and dont's of distillation control. Chem Eng Res Des (Trans I ChemE, Part A).
2007;85(A1):13-23.
26. Tyreus BD, Luyben WL. Dynamics and control of recycle systems 4. Ternary systems with one or two
recycle stream. Ind. Eng. Chem. Res. 1993, 32 (6), 1154-1162.
27. Zhang C, Vasudevan S, Rangaiah GP. Plantwide control system design and performance evaluation for
ammonia synthesis process. Ind Eng Chem Res. 2010;49(24):12538-12547.

..... .....

..... .....
....................................................................................

.....................................................

.........................................................................................................................
. . . . . ...... ................................................... .....

.....................................-..............................-........-........-........-...

.....................................-............................-........-......-......-...
...............................................................................................................
..................................................................
..................................................................
..................................... -.................................. -....... -
....................................................
䘀椀最甀爀攀..................................................................
..................................................
....................................................................................