1 Why Process Control?

The word control is encountered often in everyday conversation. We could be

looking for the missing remote control for the TV, expressing our opinion on
the issue of gun control, or trying to control our anger when faced with a frustrat-
ing situation. The Merriam-Webster Dictionary (http://www.merriam-webster.
com/) offers the following abbreviated definitions for this word most relevant to
our topic:

control transitive verb, (1) to check, test, or verify by evidence or experiments,

(2) to exercise restraining or directing influence over.
noun, (1) an act or instance of controlling; also: power or authority to guide or
manage, (2) a device or mechanism used to regulate or guide the operation of a
machine, apparatus, or system.

By controlling process systems, we simply imply influencing their operation. This

influence takes the form of guiding or regulating the operation in such a way as
to ensure a desired outcome. Perhaps you have already witnessed such instances
in the laboratory where control was part of the process operation: a heated water
bath included a control unit to maintain the desired temperature in the bath, or a
pressure regulator was used to deliver a constant air supply to the laboratory. In all
these instances, the so-called control unit is present to ensure steady operation for an
otherwise dynamic process that would exhibit significant (and sustained) deviations
from the desired target. In all these and other practical situations, it is hard not to
notice a simple fact: All industrial processes are dynamic. In other words, the pro-
cess behavior changes with time. Thus, the control engineer is called upon first and
foremost to fundamentally understand the dynamic nature of processes and then,
based on this knowledge, devise appropriate mechanisms to influence their behavior
over time.
Control engineering is no longer regarded as a narrow specialty but as an essen-
tial topic for all chemical engineers. For example, design engineers must carefully
consider the dynamic operation of all equipment and its consequences, as the plant
will never operate at steady-state, which is often the assumption that one makes to
simplify analysis and design tasks. Engineers entrusted with the operation of pro-
cessing plants must find the appropriate response to the ever-present upsets (e.g., feed
quality or quantity fluctuations, variations in the availability of various utilities), so
that the plant operates smoothly. Finally, engineers performing experiments must
control their equipment and the environment meticulously to obtain the conditions
prescribed by their experimental designs. In fact, without process control, it would

3
4 Introduction to Process Control

not be possible to operate most modern processing facilities safely, reliably, and prof-
itably, while satisfying quality and environmental standards.

1.1 HISTORICAL BACKGROUND

The history of control, especially the concept of feedback, can be traced back to
more than 2000 years ago when ancient water clocks were operated successfully
using float regulators in Alexandria (Egypt) and Baghdad (Iraq). In an issue com-
memorating the 13th World Congress of the International Federation of Automatic
Control (IFAC), Control Systems journal of the International Electronics and
Electrical Engineers (IEEE) Society published a series of articles detailing the
evolution of automatic control (Control Systems, Volume 16, No. 3, 1996), among
which the article by Stuart Bennett is an excellent reference for scientific history
enthusiasts.
Prior to the 1940s, most chemical processing plants were essentially run manually,
as the plant operators adjusted material and energy flows by manipulating rather large
valves by hand. Only the most elementary types of controllers were used to regulate a
limited number of process equipment. Many instrument companies existed, produc-
ing a variety of devices for sensing, recording, and controlling. Due to the large-scale
nature of the plants, many operators were required to keep watch on hundreds of pro-
cess variables, and they needed a strong process knowledge and expertise to respond to
unwanted trends in the best possible way. Moreover, large tanks were employed to act
as buffers or surge capacities between various units in the plant. These tanks, although
sometimes quite costly capital investments, served the function of absorbing the major-
ity of upsets by isolating parts of the process from incidents occurring upstream.
With increasing labor and equipment costs, and with the development of high-
capacity, high-performance equipment and processes in the 1940s and early 1950s,
it became uneconomical and often impossible to run plants without automatic con-
trol devices. At this stage, feedback controllers were added to the plants with little
real consideration of, or appreciation for, the underlying process dynamics. Rule-of-
thumb guidelines and process experience were the only design techniques, which
were deployed on a case-by-case basis.
By the 1960s, chemical engineers were experimenting with the new developments
in dynamic analysis and control theory, which during that decade focused on the
concept of optimal control. Most of the techniques were adapted from the work in
the aerospace and electrical engineering fields, responding to the escalation of the
arms race by dealing with missile guidance control and radar tracking systems. Foss1
and Lee and Weekman2 harshly criticized these practices as being unresponsive to
the actual needs of chemical processes and argued the uniqueness of the process
control problem in chemical industries as distinct from others.
The rapid rise in energy costs in the 1970s highlighted additional needs for
effective control applications, which are referred to as “advanced control systems.”
The design and redesign of many plants to reduce energy consumption resulted
in more complex, integrated plants that were much more interacting than before.
Consequently, the challenges facing the process control engineer have continued
to grow steadily over the years. The textbook by Stephanopoulos3 represents a
Why Process Control? 5

milestone in the conceptualization of process control as a plantwide effort, under-

scoring its fundamental link to process design.
This set the scene for the trends in the 1980s and 1990s. On one hand, theoretical
studies helped establish the credibility of the field as a rigorous discipline, although
many of the developed algorithms did not find immediate application in practice, a
notable exception being the model predictive control technology (Chapter 14), mar-
keted today in various forms by vendors like Aspen Technology and Honeywell among
others. On the other hand, control hardware went through a substantial transition from
being pneumatic, to analog/electric and to being microprocessor-based. The control
hardware that has always been designed to implement single-loop controllers and their
variations started enjoying the benefits of distributed control systems (DCS) that come
equipped with various advanced control configurations (see Chapter 21).
With the turn of the century, wireless technologies radically changed the commu-
nications paradigm, enabling not only personal communications but also how con-
trol hardware communicated with the control systems. Wireless networks provided
an ideal platform not just to transmit data but also acted as the gateways for mov-
ing data to storage and archives effortlessly. Such easy access to data is one of the
triggers of Industry 4.04 that represents the revolutionary impact of digitization and
automation on the manufacturing industries. The elements of Industry 4.0 such as
cloud computing, the Internet of Things (IoT) as well as cyber-physical systems (e.g.,
digital twins5) are already in place in most industries and this trend will continue to
shape the manufacturing practices for many years to come.
In keeping with these developments, Christofides et al.6 and Bryner7 articulate a
“Smart Plant” vision that strives for the real-time integration of process control, plant
operations, and business systems. Such integration is enabled by the aforementioned
rapid advances in cyber infrastructure and communication technologies and aims
to minimize downtime, eliminate incidents, and maximize the use of plant assets.
Exciting challenges still lay ahead for control engineers as they will participate in
and contribute to this futuristic vision as part of their day-to-day tasks.

1.2 ROLE OF CONTROL IN PROCESS INDUSTRIES

As discussed earlier, the theory and application of control has been shaped by a num-
ber of technological developments over the years. Accordingly, the role of process
control is being continuously augmented following these trends. Today, we observe
two types of roles:

1. Traditional role of process control

2. Expanded role of process control

1.2.1 TRADITIONAL ROLE OF PROCESS CONTROL

In industrial operations, control contributes to safety, environmental impact min-
imization, and process optimization, by maintaining process variables near their
desired values. A chemical reactor with a highly exothermic reaction in Figure 1.1
demonstrates a clear example of safety through control. Many input variables, such
6 Introduction to Process Control

FIGURE 1.1 A chemical reactor and its control system.

as feed composition, feed temperature, and cooling medium (coolant) temperature,

can vary, which could lead to dangerous overflow of the liquid and large tempera-
ture excursions (runaway). The control system sketched in Figure 1.1 shows that
the level is maintained near its desired value by adjusting the outlet flow rate, and
the temperature is kept near its desired value by adjusting the coolant flow rate.
At the same time, by controlling the temperature at some optimal value, the selectivity
may be maximized, thus improving process economics. This tight temperature con-
trol may also reduce the amount of byproducts, thereby easing the environmental
burden through the reduction of potential waste and further processing.
As demonstrated by the previous example, during its operation, a chemical plant
must satisfy a variety of requirements imposed by its designers along with the gen-
eral technical, economic, and social constraints, in the presence of ever-changing
external influences such as variability in supply and demand. Among such require-
ments are the following:

1. Economics: The operation of a plant (or a process unit) must conform to the
market conditions, recognizing the availability and the quality of raw mate-
rials as well as the demand for the final products. Furthermore, it should be
as economical as possible in its utilization of raw materials, energy, capital
assets, and human labor. Thus, the operating conditions must be controlled
effectively to minimize operating costs and maximize profits.
2. Production specifications: A processing plant should maintain or surpass
the production capacity that it is designed for, while delivering the final
products with consistent quality.
3. Operational constraints: The various types of equipment used in a chem-
ical plant have limitations not only as a result of their particular design
but also inherent to their operation. A centrifugal pump, for instance, can
deliver only a certain flow rate as determined by its impeller size and the
available pressure drop in the line. Control systems need to recognize and
satisfy all such operational constraints.
Why Process Control? 7

4. Safety: The safe operation of a chemical process is a primary requirement for

the well-being of plant personnel, the community at large, and the economic
viability of the company. Thus, the operating pressures, temperatures, and con-
centration of chemicals should always be maintained within allowable limits.
5. Environmental regulations: Many countries as well as local governments
have enacted laws mandating that any air, water, and ground emissions
from industrial plants must conform to specified limits in chemical compo-
sition and flow rate.

All requirements listed earlier dictate the need for continuous monitoring of the
operation of a chemical plant, with intervention when necessary, to guarantee the
satisfaction of each operational objective. This is accomplished through a rational
arrangement of equipment and instrumentation (e.g., measuring devices, valves,
controllers, and computers) along with human expertise (e.g., plant designers, plant
operators, and engineers), which altogether help deploy the control system.

1.2.2 EXPANDED ROLE OF PROCESS CONTROL

An industrial plant is a complex arrangement of various unit operations that are
designed to produce the desired product(s). An integral part of this arrangement
is the control room, where plant operators keep a watchful eye on the operation
(Figure 1.2). They use computer-based control systems for collecting process data
and adjusting the appropriate variables to maintain desired operations.
During normal operation, the plant operators should be able to rely on automatic
control systems to maintain this desired behavior. The basic unit in this effort is the
DCS that communicates with field instruments through Ethernet links and recently
through wireless networks. Figure 1.3 illustrates a computer control configuration
combining a number of control systems. Such a unit allows the plant operators to
visualize key process variables, check for alarms, and interfere with the operation,
when necessary. Such an interference takes many forms ranging from adjusting the
performance of a unit to shutting down the plant during an emergency.

FIGURE 1.2 Control room of a pulp and paper mill. (Courtesy of Visy Pulp and Paper,
Tumut, NSW, Australia.)
8 Introduction to Process Control

FIGURE 1.3 Computer control architecture and its links. (Implemented at the Process
Systems Engineering Laboratory at the Louisiana State University, Baton Rouge, LA.)

The application of computers to control the process had a profound effect on the
direction of current efforts in industrial systems control. Specifically, the computer
systems opened up new opportunities for system configuration based on (1) distrib-
uted data acquisition and control and (2) hierarchical computer control where each
computer performs selected and coordinated tasks.
As noted before, a consequence of these developments (e.g., wireless commu-
nications, cloud platforms) has been a vast broadening of the domain of what is
technologically and economically feasible to achieve in the application of comput-
ers to control industrial systems. Today, all aspects of information processing, data
gathering, process control, online optimization, even scheduling, production, and
maintenance planning functions may be included in the range of tasks to be car-
ried out by the computer control system. This has made possible the realization of
integrated control systems (ICS) in which all factors influencing plant performance
are taken into account. Chapter 21 provides a more detailed discussion of the control
system architectures and ICS concepts.
Why Process Control? 9

Control systems today have an expanded role, replacing manual manufacturing

activities with full automation, paving the way for smart manufacturing. Modern
process control is the functional integration of real-time information manage-
ment and decision making with the traditional concept of control.8

In the next section and the following chapters, we will explore several issues within
the traditional view of process control to establish a series of fundamental concepts.
Later in the book, we will return to the expanded role of control in modern manufac-
turing and build on this foundation.

1.3 OBJECTIVES OF CONTROL

A control system generally performs one or both of the following tasks:

1. Maintaining the process at the desired operating point: A process is expected

to operate at the steady-state conditions prescribed by its design. This operat-
ing point is generally the most attractive one, satisfying among others, eco-
nomic, safety, and quality objectives. Thus, if the process strays away from
this condition, substantial losses may be incurred. There may be two chief
reasons why a process may not stay at this operating point (regime).
First, the process may be unstable, implying that process variables may
not remain within their physical bounds when subjected to perturbations of
limited magnitude. Imagine a tank without any drainage being constantly
filled with water. The water level in the tank will keep rising until the tank
overflows. Such an uncontrolled behavior is intolerable in a chemical pro-
cess. Many chemical processes happen to be stable or self-regulating and
need no external intervention for their stabilization. On the other hand,
some are inherently unstable and require external manipulation for the sta-
bilization of their behavior. A chemical reactor with an exothermic reaction
is a good example in which the reactor temperature may not remain at the
desired and acceptable levels if external cooling is not sufficient. Such sys-
tems can only be operated with the aid of a control system.
Second, a process operating at steady-state experiences frequent upsets
due to various changes in its operating environment. A typical example is
when the feed stream properties (such as composition, temperature, and
flow rate) are fluctuating, thus continually pushing the process away from
its steady-state. Depending on the magnitude and period of these fluctua-
tions, process operation and performance may be significantly compro-
mised. Again, a control system would be required to provide compensation
for such disturbances and maintain the process as close as possible to the
desired operating point for as long as possible.
2. Moving the process from one operating point to another: In many process
operations, the plant personnel may want to change the operating point that
a process is at, for a variety of reasons. One example is a petroleum refinery
10 Introduction to Process Control

FIGURE 1.4 A bicycle is an inherently unstable system.

that processes different types of crude oil. The desired operating point for
a light crude would not be so desirable when a heavier crude is processed.
Or, in a polyethylene plant, quality of the product may change from time to
time depending on the customer specifications. In both of these examples,
the plant needs to adjust the operating point so that it corresponds to the
conditions best suited to the changed environment. A control system can
perform this, often complex, task in an automated manner.

Riding a bicycle (see Figure 1.4) is an attempt to stabilize an unstable system, and we
accomplish this by pedaling, steering, and leaning our body right or left. Of course,
all this is done to keep us on our chosen trajectory toward a desired destination. The
traffic or the pedestrians on the road constitute typical perturbations (disturbances)
for our ride. We, as the rider, act as the control system and may even decide to change
our destination, or stop safely for a brief rest.

1.3.1 WHAT ABOUT PERFORMANCE?

Merely surviving a fall, however, and staying on course are not the only objectives in
riding a bicycle. One should enjoy the trip as well. Therefore, an additional objective
of the control system is maintaining some level of desired performance. For process
systems, the plant engineers set the performance expectations and these typically
reflect how some key process variables evolve over time.
In control system design, good performance typically requires aggressive con-
trol actions and fast responses, while stability is generally preserved with conserva-
tive control actions and slow responses. This leads to the perennial design trade-off
between performance and stability of process control systems, as the primary goal
in advanced control design is to maximize performance while ensuring stability.

1.3.2 ECONOMIC BENEFITS OF CONTROL

The optimization of process operating conditions is performed in such a way as to
maximize an economic objective, such as the annual profit. Key factors in good
plant management are the determination of the most desirable operating targets
Why Process Control? 11

FIGURE 1.5 Optimal plant operation.

and the deployment of an effective automatic control strategy to maintain these

targets within acceptable limits. Therefore, setting the control objectives requires
a clear understanding of how the plant operates. A chemical engineer is best suited
for this task.
One can determine the most desirable plant operating conditions by first
defining the region of possible operation (feasible region) in the plane of free
(decision) variables (Figure 1.5). The feasible region is bounded by the physical
limits of these variables. Second, the plant economics are evaluated by super-
imposing the contours of increasing profits. While the optimum often lies at a
corner of the feasible region,9 in practice, the plant will be operated sufficiently
far from this point to ensure feasible operation in the presence of ever-present
plant disturbances (upsets). The control strategy aims to keep the operating con-
dition variations at a minimum and to allow the operating target to stay as close
as possible to the true (optimal) maximum profit. The control system is expected
to minimize the variations around the operating target (performance objective)
while, in turn, shrinking the tolerable operating limits. The more sophisticated
(or advanced) the control system is, the better the chances are that the plant will
operate even closer to the optimum target. This gain, quantified by the move
toward a more profitable regime, helps establishing the financial benefits of the
control system. For a more detailed discussion on optimizing control, the reader
is referred to Chapter 20.
In summary, one can conclude that good control performance has the potential
to yield substantial benefits for safe and profitable plant operation. By applying the
fundamental process control principles, the engineer will be able to design plants
and implement control strategies that can achieve the control objectives set forth by
plant operations.
12 Introduction to Process Control

1.4 SUMMARY
A good control design addresses a hierarchy of control objectives, ranging from
safety to product quality and plant profitability, which depend on the operating
objectives for the plant. These objectives are determined by both steady-state and
dynamic analysis of the plant performance. Process control aims to reduce the plant
variations and help deliver consistently high product quality and maintain operation
close to the maximum profit target.
As we have learned, good performance demands tight control of key variables.
Clearly, understanding the dynamic behavior of the process is essential in design-
ing control strategies. Only with a thorough knowledge of the process dynamics,
we can design control systems that can satisfy conflicting objectives and yield
lasting benefits.

CONTINUING PROBLEM
Let us consider a simple blending process depicted in Figure 1.6, with two feed
streams and one product stream. The feed streams contain an aqueous mixture
of a component with different compositions (mass fractions), hence requiring
a blending operation to deliver a product stream with the desired composition.
We denote the volumetric flow rate (m 3/min) of a stream with F, and x will rep-
resent the component mass fraction. The subscripts point to the individual feed
streams.
While the flow rates of the feed streams can be adjusted by process valves,
the stream compositions may vary depending on upstream processing conditions.
Moreover, the product stream is likely to have a specification on its composition
(quality), and a composition sensor (Xmeas) may be placed on this stream to monitor
its variation:

• Analyze this process and cite the specific reasons why control may be nec-
essary for its operation.
• What would be the specific benefits of a control system?

FIGURE 1.6 A blending process.

Why Process Control? 13

SOLUTION
In the blending process, the main operational concern is the quality (composition)
of the final blend stream, x. The target composition can be achieved by adjusting
the valves on the feed streams. By doing so, however, even if the composition target
is reached, the liquid level (L) in the tank may be changing. Thus, a mechanism
(a control system) becomes necessary to manage the combination of feed flow rates
not only to achieve the desired target composition of the blend but also to maintain
an acceptable level of liquid in the tank.
Due to consumer demands, the target product composition may vary and the
operation needs to respond swiftly to such demands. By maintaining the product
composition at the desired target, we minimize the production of off-specification
product that may have to be discarded or reprocessed. This strategy has clear eco-
nomic benefits. Furthermore, by maintaining a desired liquid level in the tank, we
can effectively manage the inventories, and this would have positive economic con-
sequences. We also have the added benefit of ensuring a safe and reliable operation
(i.e., the tank does not overflow or run dry).

REFERENCES
1. Foss, A.S., Critique of chemical process control theory, AIChE Journal, 19, 209, 1976.
2. Lee, W. and V.W. Weekman, Advanced control practice in the chemical industry,
AIChE Journal, 22, 27, 1976.
3. Stephanopoulos, G., Chemical Process Control—An Introduction to Theory and
Practice, Prentice-Hall, New York, 1984.
4. Gilchrist, A., Industry 4.0: The Industrial Internet of Things, Apress, 2016.
5. Qi, Q. and F. Tao, Digital twin and big data towards smart manufacturing and industry
4.0: 360 degree comparison, IEEE Access, 6, 3585, 2018.
6. Christofides, P.D., J.F. Davis, N.H. El-Farra, D. Clark, K.R.D. Harris, and J.N. Gipson,
Smart plant operations: vision, progress and challenges, AIChE Journal, 53, 2734,
2007.
7. Bryner, M., Smart manufacturing: the next revolution, Chemical Engineering Progress,
108(10), 4, 2012.
8. Erickson, K.T. and J.L. Hedrick, Plantwide Process Control, Wiley, New York, 1999.
9. Edgar, T.F., D.M. Himmelblau, and L.S. Lasdon, Optimization of Chemical Processes,
2nd edn., McGraw-Hill, New York, 2001.
 2 Definitions and
Terminology
Before tackling the main features of a control design problem, it is necessary to have
a clear understanding of some key concepts and definitions as well as the general
terminology. In this chapter, we start with the classification of process variables from
the control viewpoint. These variables are then used to analyze the characteristics of
a control system and to formulate the problems that must be solved during design.
Finally, we discuss the elements of a control design project.

2.1 CONCEPTS AND DEFINITIONS

Let us consider a system that can include a processing unit, a set of units, or a sec-
tion of the plant. Figure 2.1 depicts such a process system (a distillation column)
that interacts with its surroundings through its boundary. The engineer or the plant
operator views the process through its physical and operational characteristics. On
the other hand, the control system sees the process only through a window com-
prised of the input and output variables. Consequently, from a control viewpoint,
the variables that describe the dynamic behavior of this process can be classified
into two groups:

1. Input variables (u) represent the effect of the surroundings on the process.
2. Output variables (y) represent the effect of the process on the surroundings.

The physical process as the one illustrated in Figure 2.1 can then be represented
schematically in a simpler form using the block diagram in Figure 2.2.
Furthermore, from a practical standpoint, the input variables may or may not
be available for a deliberate action on the process, resulting in a further classifica-
tion as follows:

1. Manipulated variables (m) can be adjusted freely by a human operator or

a control system.
2. Disturbance variables (d) are not the result of an adjustment by an operator
or a control system.

According to their direct measurability, disturbance variables (or, simply, distur-

bances) themselves can be further classified as (1) measured disturbances (dm) and
(2) unmeasured disturbances (du). As examples of typical process disturbances, feed
flow rates, feed temperatures, and pressures can be easily measured using avail-
able sensor technology. Feed composition, on the other hand, is difficult to measure,

15
16 Introduction to Process Control

FIGURE 2.1 Schematic of a distillation process interacting with its surroundings.

as online composition sensors may often be expensive, unreliable, or just unavail-

able. This classification appears in Figure 2.3.

We shall see later that the unmeasured disturbances generate more difficult
control problems since their impact on the process is simply unpredictable.

The output variables, also referred to as control variables, are generally associated
with the control objectives, and are related to the process variables that indicate
product quality, process safety, and economics. They are further classified into the
following categories:

1. Measured outputs (ym) are those whose values are known on direct
measurement.
2. Unmeasured outputs (yu) are not or cannot be measured directly.

This classification is shown in Figure 2.4, and Figure 2.5 summarizes the complete
classification schematic.

FIGURE 2.2 Schematic of a process from a control viewpoint and its interactions with the
surroundings.
Definitions and Terminology 17

FIGURE 2.3 Classification of input variables.

FIGURE 2.4 Classification of output variables.

FIGURE 2.5 Schematic representation of a process with all inputs and outputs classified.
18 Introduction to Process Control

FIGURE 2.6 Classification procedure to define the problem from a control point
of view.

This procedure of process variable classification becomes very important

in practice especially when initiating a control design project, and comprises
the steps described in Figure 2.6. It starts from “the operator’s view of the
process,” proceeds with the encapsulation of the process into a block by defin-
ing the problem boundaries, and terminates at the final stage where the problem
is represented schematically according to “the controller’s view of the process.”
This final block representation provides a simpler pictorial representation of
the overall problem, the variables involved, and the possible challenges that
the operator and the engineer could face during the development of a control
strategy.
Furthermore, depending on how many output and input variables are considered
for the control problem, we can distinguish two major control structures:

1. Single-input single-output (SISO). There is a single-output variable (control

objective) and a single-input (manipulated) variable is available to influence
the process behavior.
2. Multiple-input multiple-output (MIMO). There is more than one output
variable (control objective), and more than one input (manipulated) variable
are used to affect the process behavior.

Example 2.1

Consider the stirred-tank heater shown in Figure 2.7 where a liquid is heated by
an internal coil mechanism through which a low-pressure steam is passed. Here,
F and T represent a stream flow rate and a stream temperature, respectively, and
the subscripts 0 and st refer to the inlet stream and the steam. For this process, we
have the following classification:

1. Input variables: F0, T0, Fst, and F

a. Manipulated variables: Fst, F
b. Disturbances: F0, T0 (both can be easily measured)
2. Output variables: V, and T (all can be easily measured)
Definitions and Terminology 19

FIGURE 2.7 A stirred-tank heater process for Example 2.1.

F can be considered either as an input or an output. Why?

• If there is a control valve in the effluent stream so that its flow can be
manipulated, then F may be an input variable.
• Otherwise, F is an output variable, reflecting the influence of input vari-
ables on the process.

The final classification is depicted schematically in Figure 2.8. The operational

goals for this heater process may be to maintain a constant tank level to avoid spills
and to maintain a desired liquid temperature. If we attempt to measure the liquid
holdup in the tank in addition to the effluent stream temperature and use, for
instance, the effluent flow rate and the steam flow rate as manipulated variables,
we have a MIMO control problem.

Example 2.2

Consider the distillation column in Figure 2.1. Let us assume that the feed
consists of a binary mixture to be separated into the light component at

FIGURE 2.8 Variable classification for the stirred-tank heater process.

20 Introduction to Process Control

the top of the column (distillate product) and the heavy component at the
bottom (bottoms product) using the energy provided by the reboiler and
the condenser. The feed conditions are determined by other (upstream) units
in the plant (e.g., a chemical reactor). The variables can be classified as
follows:

1. Input variables: Feed conditions (flow rate, temperature, composi-

tion); reflux flow rate; steam to reboiler (flow rate, quality); cooling
water to condenser (flow rate, quality); distillate flow rate; bottoms
flow rate
a. Possible manipulated variables: Reflux flow rate; steam flow rate;
cooling water flow rate; distillate flow rate; bottoms flow rate
b. Possible disturbances: Feed flow rate and temperature (both can be
easily measured); feed composition (may be difficult and/or expen-
sive to measure); cooling water conditions (temperature); steam con-
ditions (pressure)
2. Output variables: Temperatures along the column (trays) and in
the condenser and reboiler (can be easily measured); levels in
the reboiler and condenser (can be easily measured); distillate
and bottoms compositions (may be difficult and/or expensive to
measure)

The intermediate and final schematic representations of the process are given in
Figure 2.9 following the classification procedure.

Closely related to our discussion of the classification of variables from a control per-
spective are two key system components that provide the mechanisms for achieving
our final task (see Chapter 10 for a broader discussion), which is to control the given
process:

• Sensors provide the plant operator with information about the current state
of the system. There are a large number and a variety of process sensors,
depending on the type of variables and the information we are interested
in. Among them, we can cite a thermocouple for measuring temperature, a
gas chromatograph for measuring stream compositions, and a differential
pressure transmitter for measuring tank levels. Sensors constitute the eyes
of the control system enabling the operator to observe what is transpiring in
the process; all measured variables would naturally have sensors associated
with them.
• Actuators allow the operators to exercise their influence on the process by
adjusting the input variables. In other words, by adjusting an actuator and
changing a quantity that is delivered to the process, one would alter the
state of the process. Among the most common actuators are the control
valves and variable-speed pumps that can modify the flow rate of a process
stream. Actuators represent the muscle of the controller to help the opera-
tor drive the process in the desired operational direction; all manipulated
variables have actuators associated with them.
Definitions and Terminology 21

(a)

(b)

FIGURE 2.9 Variable classification for the distillation unit (a) and its schematic representa-
tion (b).

2.2 CONTROL DESIGN PROBLEM

The fundamental problem in control theory is to bring about a desired behavior, expressed
by a reference target (yr), of an output (control) variable (y) through the manipulation of
an input (manipulated) variable (m). The reference target is also referred to as the set-
point (ysp) and this terminology will be commonly used in later chapters.
A plant operator, being familiar with the operation of a process, can view the con-
trol variable through a display and adjust an input variable (typically a valve on a pro-
cess stream), which is expected to produce the desired outcome (Figure 2.10). This is
manual control. One can imagine many drawbacks with this effort: One operator is
responsible for a specific process (or a process unit) all the time, the operator needs
to be an expert in the operation of the process, etc.
22 Introduction to Process Control

FIGURE 2.10 Manual control.

To solve this problem more effectively, we can, in principle, declare that there is a
special mechanism, called the control mechanism, yet to be determined. The control
mechanism will adjust the manipulated variable in such a way as to cause the output
variable to follow the specified reference value as dictated by the control objective.
Conceptually, this is depicted in Figure 2.11 and referred to as open-loop control.
Let us assume that a perfect mathematical representation (model) of the process
is available, which exactly describes how y(t) changes in response to m(t). Then, one
can ask: Can this model be used to determine m(t), if the desired (reference) output,
yr(t), is specified? The control mechanism, then, can simply become this inverse
model. Indeed, this leads to a perfect control design as we would compute precisely
what value of m(t) to send to the process so that y(t) follows yr(t) exactly. While this
appears to be an intuitive solution to our control design problem, there may be practi-
cal difficulties in implementing it:

1. In the first place, the assumption of the availability of a perfect process

model is never true.
2. It may be computationally difficult or impossible to actually invert the
model, that is, solve for m(t) as a function of yr(t).
3. This approach, which we refer to as open-loop, does not allow the controller
to “find out” the actual state of the process if and when it differs from the
model prediction.

FIGURE 2.11 Open-loop control.

Definitions and Terminology 23

FIGURE 2.12 The feedforward control scheme.

The latter point becomes crucial when the control mechanism does not have perfect
knowledge of the disturbances affecting the process. In reality, the process is subject
to disturbances, and the control mechanism needs information regarding the current
value of the disturbance. This leads to an alternative control structure, known as the
feedforward control, if we assume that this disturbance is measured (Figure 2.12).
As one can observe, the control mechanism receives information about changes in
the disturbances and helps produce an anticipatory control action m(t) to counteract
the effect of the disturbance on the process.
In many practical cases, however, the disturbances will (or can) not be measured;
hence, our knowledge of the process will be imperfect. Then, it becomes impera-
tive to supply current process information to the control mechanism so that it keeps
track of the process behavior. This is accomplished through the measurement of the
output (control) variables as shown in Figure 2.13. We observe that, in this case, the
information obtained from the direct measurement of the output of the process is fed
back to the control mechanism. In other words, the mechanism uses the knowledge
of the status of the plant to generate the requisite control signal. This is the concept
behind feedback control.
Based on the preceding discussion, we identify four possible configurations:

Case 0: A human operator implements the control action. This is a manual

control scheme.
Case 1: The control mechanism acts without current information about the
status of the process. This is an open-loop control scheme.

FIGURE 2.13 The feedback control scheme.

24 Introduction to Process Control

Case 2: The control mechanism anticipates the effect of the disturbances,

producing a corrective action. This is an anticipatory control action or a
feedforward control scheme.
Case 3: The control mechanism acts using the information fed back from the
measurements. This is a feedback control scheme.

Example 2.3

Consider the blending process as detailed in the “Continuing Problem” at the end
of Chapter 1. A number of control configurations are possible for this system as
described previously (Figure 2.14).

• Manual: The operator monitors the desired blend composition and,

being familiar with the operation of the blending process, makes appro-
priate adjustments to the feed flow rates to achieve the desired blend.

FIGURE 2.14 Alternative control configurations for the blending process (a) open-loop,
(b) feedforward, (c) feedback, and (d) combined feedforward-feedback.
Definitions and Terminology 25

• Feedforward: The control mechanism receives information from the

measurement of the flow rate on one of the feed streams and compen-
sates for this variation by adjusting the valve opening on the other feed
stream.
• Feedback: The control mechanism receives information from the mea-
surement of the product composition and adjusts the valve opening on
one of the feed streams.
• Combination: What if we combined the last two control mechanisms?
In this schematic, the valve opening on one of the feed streams would
be adjusted based on measurements from the flow rate variations in the
other feed stream and the product composition.

Naturally, there are many more possible alternatives as we shall see later in this
book and the cases, just discussed, are the most typical ones encountered in
practice.

In summary, we observe that control design consists of a series of decision-making

steps leading to a comprehensive design strategy.

2.3 CONTROL SYSTEM DESIGN

An engineer approaches any design problem in a systematic manner. Accordingly,
there is a series of steps to follow during the design of a control system to facilitate
the decision-making process of the control engineer. They are as follows:

1. Define control objectives: This step is about defining the operational objec-
tives that a control system is called upon to achieve. The key objective is to
maintain the process output variables as close as possible to their targets.
By ensuring stability, eliminating disturbances, and optimizing the eco-
nomic performance of the process, the control system strives to achieve this
objective. Naturally, this generic objective statement needs to be translated
into specific control objectives for the process in question. This is the task
of the control engineer.
2. Select measured variables: Second, we need some means to monitor the
performance of the process and observe how it may respond to actions by
the control system. This is accomplished by measuring the values of certain
process variables (e.g., temperature, pressure, compositions). It is evident
that we shall directly monitor the variables that represent (or indicators of)
the control objectives and this is done whenever possible (we discuss cases
later showing how a control system is designed when direct measurements
are not possible).
3. Select manipulated variables: Once the control objectives are specified and
the measurements are identified, the next question is how we can cause a
change in the process. In other words, what are the variables to be manip-
ulated to achieve the control objectives? Usually, we have many options
in this selection, and our decision may affect the quality of the process
response, that is, the control performance.
26 Introduction to Process Control

4. Select control configuration: The control configuration is the informa-

tion structure used to match the available measurements with the available
manipulated variables. Normally, we will have many possible configura-
tions (feedback, feedforward, and their combinations, SISO, MIMO, etc.).
What is best for a given problem is a critical question for the quality of the
control system, and is problem specific.
5. Tune the controller: The controller is the active element that receives the
information from the measurements and takes appropriate actions to adjust
the values of the manipulated variables. The question here is how the infor-
mation taken from the measurements is used to adjust the values of the
manipulated variables and the answer constitutes the control law, which is
executed automatically by the controller. As we shall see in later chapters,
tuning implies determination of the parameters of the controller.

Example 2.4

For the stirred-tank heater process in Example 2.1, the steps are as follows:

1. Control objective: Maintain the volume (level) and the temperature of

the liquid in the tank at their desired (target) values in the presence of
variations in the upstream conditions (disturbances), namely, fluctuations
in the inlet stream temperature and flow rate.
2. Control variables: Our first attempt is to install measuring devices to
monitor the temperature of the liquid in the tank, T (remember this is the
same temperature as the exit stream temperature due to the perfect mix-
ing assumption), and the liquid level, h. This is accomplished by using a
thermocouple (for T) and a differential pressure cell (for h).
3. Manipulated variables: We can choose to manipulate the outlet flow
rate, F, and the steam flow rate, Fst. Based on our engineering intuition,
we can argue that the steam flow rate will only affect the tank tempera-
ture and by varying the outlet flow rate, we can influence the liquid level
in the tank.
4. Control configuration: It is possible to choose either a feedback control
scheme to deal with the disturbances (inlet stream flow rate and tem-
perature) or even a feedforward scheme, since these disturbances can
be easily measured. With two control variables and two manipulated
variables, this is a MIMO control system.
5. Controller tuning: Depending on the type of function that we suggest
relating the control variables to the manipulated variables, we have to
determine the parameters of this function. We have to propose relevant
criteria for judging the acceptability of this set of parameters. Such crite-
ria typically involve measures of performance for the controlled dynamic
behavior of the process.

This systematic approach to control system design needs to be placed in perspective

with respect to the inception, execution, and implementation of a plantwide design
project.
Definitions and Terminology 27

2.4 CONTROL DESIGN PROJECT

As discussed in Chapter 1, the computer control systems are an integral part of enter-
prise management as they complement other plantwide activities such as informa-
tion processing, data gathering, online optimization, and production planning and
scheduling. This makes the task of the control engineer much more complex than
what it has been in the past, which mostly centered around isolated instrumentation
activities. In their new role, the control engineers view the needs of the plant from a
control standpoint to understand the information management tasks associated with
these needs and to make sure that they are consistent with the business goals of the
company. Therefore, the control design project is a team effort that moves through a
series of stages before reaching its conclusion.
Project teams are brought together to perform either a grassroots design project
or a retrofit design project for the plant. While the former aims to build a facility
from the ground up, the latter focuses on minor or major improvements in an exist-
ing facility. Such a capital project is managed through an integrated approach that
takes into account the plant personnel, process technology, as well as the control
and information technology.1 Being part of the latter area, the control engineer
is expected to maintain a broad perspective in achieving the goals of the control
design project.
The plantwide control system project consists of the following activities as dis-
cussed in detail by Erickson and Hedrick1:

• Preliminary engineering
• Detailed engineering
• Implementation
• Installation
• Commissioning
• Final production start-up and turnover
• Training

2.4.1 PRELIMINARY ENGINEERING

The control design project should be considered as part of the overall plant design
project from the beginning. Ideally, operability of a plant should be one of the objec-
tives during the conceptual design of a process along with profitability, feasibility,
safety, and others. Such an integrated approach would reduce or eliminate costly
overdesign factors and operational bottlenecks and result in a more efficient and flex-
ible plant operation. While the control project traditionally has been started after the
plant design is finalized, this trend is changing.
The control engineer typically works with a process flow diagram (PFD) gener-
ated during the conceptual plant design project. The PFD associated with a process
may show less or greater detail depending on how far the design project has pro-
gressed. Generally, the PFD will include major process equipment; their connec-
tions, depending on the design stage, may show material and energy balances and
28 Introduction to Process Control

preliminary sizing information. For the control engineer, such a PFD can be the
basis for the following activities:

• To develop the process operation description

• To develop the control concept
• To define the preliminary automation and control strategy
• To define the preliminary system architecture
• To prepare the preliminary control technology budget

These activities help establish the boundaries of the process area, identify control
objectives associated with units (as described in Section 2.3), define the extent of
automation and the level and type of control system required, decide on the hardware
and software platforms to be used for implementation, and help create a preliminary
budget. In most cases, the control engineer works with control system vendors to
evaluate the capabilities of available technologies and their relevance to the process
control problems articulated at this stage.

2.4.2 DETAILED ENGINEERING

Once the preliminary control design project receives approval, more detailed design
activities are conducted. These activities typically result in the following products
(deliverables):

• Specification of measurement and final control functions

• Specification of safety instrumented system functions
• Specification of discrete and regulatory control functions
• Specification of procedural control functions
• Specification on process information data models
• Specification of run management functions
• Specification of user-interface functions

These specifications (associated with performance, size, reliability, etc.) guide con-
trol technology development in all areas of hardware, software, and programming.

2.4.3 IMPLEMENTATION
At this stage, the hardware for measurement and final control elements is procured.
The software associated with these elements is also obtained and tested to ensure
that they meet design specifications. Implementation is not complete until the system
is verified by these tests.

2.4.4 INSTALLATION
After the hardware and associated software are successfully tested, they are ready
for installation at the site. This involves establishing wired or wireless interconnec-
tions and their verification. The control system is set up, and communications over
the local (and if necessary, global) network are tested.
Definitions and Terminology 29

2.4.5 COMMISSIONING
This is the stage where each plant system is brought online in a systematic man-
ner and the associated technology is verified to satisfy the overall operational
goals of the process. Initial control parameters can be changed (tuned) and vari-
ous control settings can be experimented with to evaluate their impact on process
performance.

2.4.6 FIRST PRODUCTION START-UP AND TURNOVER

Once the commissioning step is completed, the plant is ready for its first production.
The control engineer (along with the design engineers) oversees the start-up and, if
this is satisfactory, turns over the plant to the plant personnel (operators). The control
design project terminates at this point.

2.4.7 TRAINING
Training is a critical activity that continues throughout the operational lifetime
of the plant. It starts after the turnover and may focus on system-specific details
as well as emergency handling activities. Training is a valuable activity espe-
cially for new personnel and can be customized according to their position and
duties.

2.5 SUMMARY
In this chapter, we introduced the concepts associated with a process control prob-
lem, starting with the classification of variables. The control design problem involves
the definition of feedback control and how it is different from open-loop and feed-
forward control configurations. The control system design starts with the definition
of the control objective and the selection of control and manipulated variables that
target this objective. It concludes with the selection of the control configuration and
the determination of parameters in the control law. The control system design is part
of the control design project that is, in turn, part of the capital plant project. With
this perspective, this chapter also discusses the elements of the control design project
from the preliminary design all the way to the commissioning of the control system
and the start-up and turnover of the plant.

CONTINUING PROBLEM
A blending process was introduced in Chapter 1 (Figure 1.6). For this process,

• Classify all process variables in terms of the categories discussed in this

chapter.
• Define the control objectives.
• Propose a set of control and manipulated variables and a control configura-
tion for this process.
30 Introduction to Process Control

FIGURE 2.15 Classification of variables for the blending process.

SOLUTION
Classification of Variables (Figure 2.15)
1. Input variables: F1, F2, x1, x2
a. Possible manipulated variables: F1, F2
b. Possible unmeasured disturbances: x1, x2
2. Output variables: x, V
a. Possible measured variables: x, V

If the cross-sectional area of the tank is denoted by A, and the height of the liquid in
the tank is h, the liquid volume in the tank can be expressed as

V = Ah

We note that the volume changes as the level (height) of the liquid in the tank changes.
Furthermore, the exit flow rate F also varies with the level as the flow depends on the
static liquid height in the tank. Thus, the actual measured variables to be considered
are h and x.

Control Objectives
The control objectives are to produce a product of certain mass fraction x and main-
tain it at its desired target value, and to maintain a constant liquid level h in the tank.

Control Configuration
The level control can be accomplished by measuring the liquid level and manipulat-
ing one of the input flow rates. The composition control, on the other hand, would
require measuring the blend composition and manipulating the other feed flow rate.

REFERENCE
1. Erickson, K.T. and J.L. Hedrick, Plantwide Process Control, Wiley, New York, 1999.