Chapter 1

Introduction to Process Control

CHAPTER CONTENTS
1.1 Representative Process Control Problems
1.1.1 Continuous Processes
1.1.2 Batch and Semi-Batch Processes
1.2 Illustrative Example-A Blending Process
1.3 Classification of Process Control Strategies
1.3.1 Process Control Diagrams
1.4 A More Complicated Example-A Distillation Column
1.5 The Hierarchy of Process Control Activities
1.6 An Overview of Control System Design
Summary

In recent years the performance requirements for emphasize steady-state and equilibrium conditions in such
process plants have become increasingly difficult to sat- courses as material and energy balances, thermodynam-
isfy. Stronger competition, tougher environmental and ics, and transport phenomena. But process dynamics are
safety regulations, and rapidly changing economic con- also very important. Transient operation occurs during
ditions have been key factors in tightening product important situations such as start-ups and shutdowns,
quality specifications. A further complication is that unusual process disturbances, and planned transitions
modern plants have become more difficult to operate from one product grade to another. Consequently, the
because of the trend toward complex and highly inte- first part of this book is concerned with process dynamics.
grated processes. For such plants, it is difficult to pre- The primary objective of process control is to main-
vent disturbances from propagating from one unit to tain a process at the desired operating conditions, safely
other interconnected units. and efficiently, while satisfying environmental and
In view of the increased emphasis placed on safe, product quality requirements. The subject of process
efficient plant operation, it is only natural that the subject control is concerned with how to achieve these goals.
of process control has become increasingly important in In large-scale, integrated processing plants such as oil
recent years. Without computer-based process control refineries or ethylene plants, thousands of process vari-
systems it would be impossible to operate modern ables such as compositions, temperatures, and pressures
plants safely and profitably while satisfying product are measured and must be controlled. Fortunately,
quality and environmental requirements. Thus, it is im- large numbers of process variables (mainly flow rates)
portant for chemical engineers to have an understand- can usually be manipulated for this purpose. Feedback
ing of both the theory and practice of process control. control systems compare measurements with their de-
The two main subjects of this book are process sired values and then adjust the manipulated variables
dynamics and process control. The term process dynamics accordingly.
refers to unsteady-state (or transient) process behavior. As an introduction to the subject, we consider repre-
By contrast, most of the chemical engineering curricula sentative process control problems in several industries.
1
2 Chapter 1 Introduction to Process Control

1.1 REPRESENTATIVE PROCESS petroleum fractions by the heat transferred

CONTROL PROBLEMS from a burning fuel/air mixture. The furnace
temperature and amount of excess air in the flue
The foundation of process control is process under-
gas can be controlled by manipulating the fuel
standing. Thus, we begin this section with a basic question:
flow rate and the fuel/air ratio. The crude oil
what is a process? For our purposes, a brief definition is
composition and the heating quality of the fuel
appropriate: are common disturbance variables.
Process: The conversion of feed materials to (d) Multicomponent distillation column. Many dif-
products using chemical and physical operations. In ferent control objectives can be formulated for
practice, the term process tends to be used for both distillation columns. For example, the distillate
the processing operation and the processing composition can be controlled by adjusting the
equipment. reflux flow rate or the distillate flow rate. If
Note that this definition applies to three types of com- the composition cannot be measured on-line, a
mon processes: continuous, batch, and semi-batch. Next, tray temperature near the top of the column can
we consider representative processes and briefly summa- be controlled instead. If the feed stream is sup-
rize key control issues. plied by an upstream process, the feed conditions
will be disturbance variables.

1.1.1 Continuous Processes For each of these four examples, the process control
problem has been characterized by identifying three
Four continuous processes are shown schematically in important types of process variables.
Figure 1.1:
• Controlled variables (CVs): The process variables
(a) Tubular heat exchanger. A process fluid on that are controlled. The desired value of a controlled
the tube side is cooled by cooling water on the variable is referred to as its set point.
shell side. Typically, the exit temperature of
• Manipulated variables (MVs): The process variables
the process fluid is controlled by manipulating
that can be adjusted in order to keep the controlled
the cooling water flow rate. Variations in the variables at or near their set points. Typically, the
inlet temperatures and the process fluid flow manipulated variables are flow rates.
rate affect the heat exchanger operation. Con-
sequently, these variables are considered to be • Disturbance variables (DVs): Process variables
disturbance variables. that affect the controlled variables but cannot be
manipulated. Disturbances generally are related
(b) Continuous stirred-tank reactor (CSTR). If the
to changes in the operating environment of the
reaction is highly exothermic, it is necessary to process: for example, its feed conditions or ambi-
control the reactor temperature by manipulating ent temperature. Some disturbance variables can
the flow rate of coolant in a jacket or cooling be measured on-line, but many cannot such as the
coil. The feed conditions (composition, flow rate,
crude oil composition for Process (c), a thermal
and temperature) can be manipulated variables
cracking furnace.
or disturbance variables.
(c) Thermal cracking furnace. Crude oil is broken The specification of CVs, MVs, and DVs is a critical step
down ("cracked") into a number of lighter in developing a control system. The selections should

Combustion
Reactants products

Coolin~
Disti Ilate

medium , Products I I
Process ,
fluid--,
I Cracked
products

Cooling
medium
I 0 D Crude
oil
Feed

L
I
Coolant
out
Bottoms

(a) Heat (b) Chemical (c) Cracking (d) Distillation

exchanger reactor furnace column

Figure 1.1 Some typical continuous processes.

 1.2 Illustrative Example-A Blending Process 3

Etching 1 1m pure
gases 1 blood

....c
Plasma :.3
"'c
Q. Dialysis
medium
"'
E
:::J
I

Wafer 1 Purified
blood

(e) Semi-batch ({)Wood chip (g) Plasma (h) Kidney

reactor digester etcher dialysis unit

Figure 1.2 Some typical processes whose operation is noncontinuous.

be based on process knowledge, experience, and control (g) Plasma etcher in semiconductor processing. A
objectives. single wafer containing hundreds of printed cir-
cuits is subjected to a mixture of etching gases
under conditions suitable to establish and main-
1.1.2 Batch and Semi-Batch Processes
tain a plasma (a high voltage applied at high
Batch and semi-batch processes are used in many temperature and extremely low pressure). The
process industries, including microelectronics, pharma- unwanted material on a layer of a microelec-
ceuticals, specialty chemicals, and fermentation. Batch tronics circuit is selectively removed by chemical
and semi-batch processes provide needed flexibility for reactions. The temperature, pressure, and flow
multiproduct plants, especially when products change rates of etching gases to the reactor are con-
frequently and production quantities are small. Fig- trolled by adjusting electrical heaters and control
ure 1.2 shows four representative batch and semi-batch valves.
processes: (h) Kidney dialysis unit. This medical equipment is
(e) Batch or semi-batch reactor. An initial charge used to remove waste products from the blood
of reactants is brought up to reaction conditions, of human patients whose own kidneys are failing
and the reactions are allowed to proceed for a or have failed. The blood flow rate is maintained
specified period of time or until a specified by a pump, and "ambient conditions," such as
conversion is obtained. Batch and semi-batch temperature in the unit, are controlled by ad-
reactors are used routinely in specialty chemical justing a flow rate. The dialysis is continued long
plants, polymerization plants (where a reaction enough to reduce waste concentrations to accept-
byproduct typically is removed during the reac- able levels.
tion), and in pharmaceutical and other biopro- Next, we consider an illustrative example in more detail.
cessing facilities (where a feed stream, e.g.,
glucose, is fed into the reactor during a portion
of the cycle to feed a living organism, such as a 1.2 ILLUSTRATIVE EXAMPLE-A
yeast or protein). Typically, the reactor temper- BLENDING PROCESS
ature is controlled by manipulating a coolant A simple blending process is used to introduce some
flow rate. The end-point (final) concentration of important issues in control system design. Blending op-
the batch can be controlled by adjusting the de- erations are commonly used in many industries to en-
sired temperature, the flow of reactants (for sure that final products meet customer specifications.
semi-batch operation), or the cycle time. A continuous, stirred-tank blending system is
(f) Batch digester in a pulp mill. Both continuous shown in Fig. 1.3. The control objective is to blend
and semi-batch digesters are used in paper man- the two inlet streams to produce an outlet stream that
ufacturing to break down wood chips in order to has the desired composition. Stream 1 is a mixture of
extract the cellulosic fibers. The end point of the two chemical species, A and B. We assume that its
chemical reaction is indicated by the kappa mass flow rate w1 is constant, but the mass fraction of
number, a measure of lignin content. It is con- A, x1, varies with time. Stream 2 consists of pure A
trolled to a desired value by adjusting the di- and thus x 2 = 1. The mass flow rate of Stream 2, w 2 ,
gester temperature, pressure, and/or cycle time. can be manipulated using a control valve. The mass
4 Chapter 1 Introduction to Process Control

Control valve
Mixture of A and B Pure A
Xl--------., x2 = 1
wl '¥ w2

._________Overflow Iine

Figure 1.3 Stirred-tank blending system.

fraction of A in the exit stream is denoted by x and Next we consider several strategies for reducing the
the desired value (set point) by Xsp. Thus for this effects of x1 disturbances on x.
control problem, the controlled variable is x, the
manipulated variable is w2 , and the disturbance vari- Method I. Measure x and adjust Wz. It is reasonable to
able is x1. measure controlled variable x and then adjust w2 ac-
Next we consider two questions. cordingly. For example, if x is too high, w2 should be
reduced; if x is too low, w2 should be increased. This
Design Question. If the nominal value of x1 is :X1, control strategy could be implemented by a person
what nominal flow rate w2 is required to produce the (manual control). However, it would normally be more
desired outlet concentration, Xsp? convenient and economical to automate this simple
To answer this question, we consider the steady-state task (automatic control).
material balances: Method 1 can be implemented as a simple control
Overall balance: algorithm (or control law),

0 = w1 + Wz w (1-1) Wz(t) = Wz + Kc[Xsp - x(t)] (1-4)

Component A balance: where Kc is a constant called the controller gain. The

symbols, w2(t) and x(t), indicate that w2 and x change
wx (1-2) with time. Equation 1-4 is an example of proportional
control, because the change in the flow rate, w2 (t) - w2 ,
The overbar over a symbol denotes its nominal steady-
is proportional to the deviation from the set point,
state value, for example, the value used in the process
Xsp- x(t). Consequently, a large deviation from set
design. According to the process description, :X2 = 1
point produces a large corrective action, while a small
and :X = Xsp· Solving Eq. 1-1 for w, substituting these
deviation results in a small corrective action. Note that
values into Eq. 1-2, and rearranging gives:
we require Kc to be positive because w2 must increase
- Xsp - X1 when x decreases, and vice versa. However, in other
Wz = w1 1 - Xsp (1-3) control applications, negative values of Kc are appro-
priate, as discussed in Chapter 8.
Equation 1-3 is the design equation for the blending A schematic diagram of Method 1 is shown in Fig. 1.4.
system. If our assumptions are correct and if x 1 = x1 , The outlet concentration is measured and transmitted to
then this value of w 2 will produce the desired result, the controller as an electrical signal. (Electrical signals
x = Xsp· But what happens if conditions change? are shown as dashed lines in Fig. 1.4.) The controller ex-
Control Question. Suppose that inlet concentration ecutes the control law and sends the calculated value of
x 1 varies with time. How can we ensure that the outlet w2 to the control valve as an electrical signal. The con-
composition x remains at or near its desired value, trol valve opens or closes accordingly. In Chapters 8 and
Xsp? 9 we consider process instrumentation and control hard-
ware in more detail.
As a specific example, assume that x1 increases to a
constant value that is larger than its nominal value, :X1. Method 2. Measure x1, adjust w2 . As an alternative to
It is clear that the outlet composition will also increase Method 1, we could measure disturbance variable x 1
due to the increase in inlet composition. Consequently, and adjust w2 accordingly. Thus, if x 1 > x1 , we would
at this new steady state, x > Xsp· decrease Wz so that Wz < Wz. If x1 < :X1, we would
 1.3 Classification of Process Control Strategies 5

1.3 CLASSIFICATION OF PROCESS

----Electrical signal CONTROL STRATEGIES
Next, we will classify the four blending control strate-
gies of the previous section and discuss their relative
~~-----,t advantages and disadvantages. Method 1 is an example
of a feedback control strategy. The distinguishing fea-
ture of feedback control is that the controlled variable
is measured, and that the measurement is used to ad-
Composition just the manipulated variable. For feedback control,
analyzer/transmitter the disturbance variable is not measured.
It is important to make a distinction between nega-
tive feedback and positive feedback. In the engineering
literature, negative feedback refers to the desirable sit-
uation in which the corrective action taken by the con-
troller forces the controlled variable toward the set
Figure 1.4 Blending system and Control Method 1. point. On the other hand, when positive feedback oc-
curs, the controller makes things worse by forcing the
increase w 2 • A control law based on Method 2 can be controlled variable farther away from the set point.
derived from Eq. 1-3 by replacing :X1 with x1(t) and w2 For example, in the blending control problem, positive
with w2 (t): feedback takes place if Kc < 0, because w 2 will in-
crease when x increases. 1 Clearly, it is of paramount
_ Xsp x1(t) importance to ensure that a feedback control system
wz(t) = w1 -----,-1- - - (1-5)
Xsp incorporates negative feedback rather than positive
The schematic diagram for Method 2 is shown in Fig. feedback.
1.5. Because Eq. 1-3 is valid only for steady-state condi- An important advantage of feedback control is that
tions, it is not clear just how effective Method 2 will be corrective action occurs regardless of the source of
during the transient conditions that occur after an x 1 the disturbance. For example, in the blending process,
disturbance. the feedback control law in (1-4) can accommodate
disturbances in w1, as well as x1. Its ability to handle
Method 3. Measure x 1 and x, adjust w 2 . This approach disturbances of unknown origin is a major reason why
is a combination of Methods 1 and 2. feedback control is the dominant process control strat-
Method 4. Use a larger tank. If a larger tank is used, egy. Another important advantage is that feedback con-
fluctuations in x 1 will tend to be damped out as a result trol reduces the sensitivity of the controlled variable to
of the larger volume of liquid. However, increasing unmeasured disturbances and process changes. However,
tank size is an expensive solution due to the increased feedback control does have a fundamental limitation: no
capital cost. corrective action is taken until after the disturbance has
upset the process, that is, until after the controlled vari-
able deviates from the set point. This shortcoming is evi-
Composition dent from the control law of (1-4).
controller Method 2 is an example of a feedforward control strat-
,-----~-------1 egy. The distinguishing feature of feedforward control is
Composition
I I
I that the disturbance variable is measured, but the con-
I
analyzer/transmitter
Control
trolled variable is not. The important advantage of feed-
xl valve x2 = 1 forward control is that corrective action is taken before
wl------~-, w2 the controlled variable deviates from the set point. Ide-
ally, the corrective action will cancel the effects of the
disturbance so that the controlled variable is not affected
by the disturbance. Although ideal cancelation is gener-
ally not possible, feedforward control can significantly

1Note that social scientists use the terms negative feedback and
positive feedback in a very different way. For example, they would say
that teachers provide "positive feedback" when they compliment
students who correctly do assignments. Criticism of a poor performance
Figure 1.5 Blending system and Control Method 2. would be an example of "negative feedback."
6 Chapter 1 Introduction to Process Control

Table 1.1 Concentration Control Strategies for the Blending 1.3.1 Process Control Diagrams
System
Next we consider the equipment that is used to imple-
Measured Manipulated ment control strategies. For the stirred-tank mixing
Method Variable Variable Category system under feedback control in Fig. 1.4, the exit con-
1 X wz FB centration xis controlled and the flow rate Wz of pure
2 X] wz FF species A is adjusted using proportional control. To
3 x1 andx wz FF/FB consider how this feedback control strategy could be im-
4 Design change plemented, a block diagram for the stirred-tank control
system is shown in Fig. 1.6. Operation of the concentra-
FB = feedback control; FF = feedforward control; FF/FB =
tion control system can be summarized for the key
feedforward control and feedback control.
hardware components as follows:
reduce the effects of measured disturbances, as discussed 1. Analyzer and transmitter: The tank exit concentra-
in Chapter 15. tion is measured by means of an instrument that
Feedforward control has three significant disadvan- generates a corresponding milliampere (mA)-level
tages: (i) the disturbance variable must be measured (or signal. This time-varying signal is then sent to the
accurately estimated), (ii) no corrective action is taken controller.
for unmeasured disturbances, and (iii) a process model 2. Feedback controller: The controller performs three
is required. For example, the feedforward control strat- distinct calculations. First, it converts the actual set
egy for the blending system (Method 2) does not take point Xsp into an equivalent internal signal Xsp·
any corrective action for unmeasured w1 disturbances. Second, it calculates an error signal e(t) by sub-
In principle, we could deal with this situation by mea- tracting the measured value Xm(t) from the set point
suring both x1 and w1 and then adjusting Wz accordingly. isp' that is, e(t) = Xsp - Xm(t). Third, c~ntroller out-
However, in industrial applications it is generally un- put p(t) is calculated from the proportiOnal control
economical to attempt to measure all potential distur- law similar to Eq. 1-4.
bances. A more practical approach is to use a combined 3. Control valve: The controller output p(t) in this case
feedforward-feedback control system, in which feed- is a DC current signal that is sent to the control
back control provides corrective action for unmeasured valve to adjust the valve stem position, which in
disturbances, while feedforward control reacts to elimi- turn affects flow rate w 2(t). Because many control
nate measured disturbances before the controlled vari- valves are pneumatic, i.e., are operated by air pres-
able is upset. Consequently, in industrial applications sure, the controller output signal may have to be
feedforward control is normally used in combination converted to an equivalent air pressure signal capa-
with feedback control. This approach is illustrated by ble of adjusting the valve position. For simplicity,
Method 3, a combined feedforward-feedback control we do not show such a transducer in this diagram.
strategy because both x and x1 are measured.
Finally, Method 4 consists of a process design change The block diagram in Fig. 1.6 provides a convenient start-
and thus is not really a control strategy. The four strate- ing point for analyzing process control problems. The
gies for the stirred-tank blending system are summarized physical units for each input and output signal are also
in Table 1.1.
Calculations performed
by controller
I
I
I x1 [mass
I
I fraction]
I Comparator
Xsp : Analyzer~ Feedback p(t) Control w2(t) Stirred x(t)
_[_m_a-ss---HIo-l calibration ~ controller [mAl valve I [kg/~ tank [mass
fraction]: '-------~ fraction]
I
I
I
I
I
I
I

xm(t) Analyzer x(t)

Figure 1.6 Block diagram for composition L _ ____:.:.::..__ _ __, (sensor) and ~---___::..::.,: _ _ _ ___J

feedback control system in Fig. 1.4. [mAl transmitter

 1.4 A More Complicated Example-A Distillation Column 7

shown. Note that the schematic diagram in Fig. 1.4. medium lines. The feed stream is assumed to come from
shows the physical connections between the compo- an upstream unit. Thus, the feed flow rate cannot be
nents of the control system, while the block diagram manipulated, but it can be measured and used for feed-
shows the flow ofinformation within the control system. forward control.
The block labeled "control valve" has p(t) as its input A conventional multiloop control strategy for this
signal and w 2 (t) as its output signal, which illustrates distillation column would consist of five feedback con-
that the signals on a block diagram can represent either trol loops. Each control loop uses a single manipulated
a physical variable such as w 2 (t) or an instrument signal variable to control a single controlled variable. But
such as p(t). how should the controlled and manipulated variables
Each component in Fig. 1.6 exhibits behavior that can be paired? The total number of different multiloop
be described by a differential or algebraic equation. One control configurations that could be considered is 5!, or
of the tasks facing a control engineer is to develop suit- 120. Many of these control configurations are impracti-
able mathematical descriptions for each block; the de- cal or unworkable, such as any configuration that at-
velopment and analysis of such dynamic mathematical tempts to control the base level hs by manipulating
models are considered in Chapters 2-7. distillate flow D or condenser heat duty Qv. However,
Other elements in the block diagram (Fig. 1.6) are dis- even after the infeasible control configurations are
cussed in detail in future chapters. Sensors and control eliminated, there are still many reasonable configura-
valves are presented in Chapter 9, and the feedback con- tions left. Thus, there is a need for systematic techniques
troller is covered in Chapter 8. that can identify the most promising configurations. For-
tunately, such tools are available; these are discussed in
Chapter 18.
1.4 A MORE COMPLICATED EXAMPLE-
In control applications, for which conventional multi-
A DISTILLATION COLUMN
loop control systems are not satisfactory, an alternative
The blending control system in the previous section is approach, multivariable control, can be advantageous.
quite simple, because there is only one controlled vari- In multivariable control, each manipulated variable is
able and one manipulated variable. For most practical adjusted based on the measurements of all the con-
applications, there are multiple controlled variables and trolled variables rather than only a single controlled
multiple manipulated variables. As a representative variable, as in multiloop control. The adjustments are
example, we consider the distillation column in Fig. 1.7, based on a dynamic model of the process that indicates
which has five controlled variables and five manipulated how the manipulated variables affect the controlled
variables. The controlled variables are product compo- variables. Consequently, the performance of multivari-
sitions, xv and xs, column pressure, P, and the liquid able control, or any model-based control technique, will
levels in the reflux drum and column base, hv and hs. depend heavily on the accuracy of the process model. A
The five manipulated variables are product flow rates, specific type of multivariable control, model predictive
D and B, reflux flow, R, and the heat duties for the con- control, has had a major impact on industrial practice,
denser and reboiler, Qv and Qs. The heat duties are ad- as discussed in Chapter 20.
justed via the control valves on the coolant and heating

Feed ---------;~

AT: analyzer/transmitter
LT: level transmitter
PT: pressure transmitter

Bottoms Figure 1.7 Controlled and

~--~~~~f~~ B manipulated variables for a
XB typical distillation column.
8 Chapter 1 Introduction to Process Control

1.5 THE HIERARCHY OF PROCESS measurement and actuation functions are an indis-
CONTROL ACTIVITIES pensable part of any control system.
As mentioned earlier, the chief objective of process
control is to maintain a process at the desired operating Safety and Environmental/Equipment Protection
conditions, safely and efficiently, while satisfying envi- (Level2)
ronmental and product quality requirements. So far, we The Level2 functions play a critical role by ensuring that
have emphasized one process control activity, keeping the process is operating safely and satisfies environmen-
controlled variables at specified set points. But there tal regulations. As discussed in Chapter 10, process
are other important activities that we will now briefly safety relies on the principle of multiple protection layers
describe. that involve groupings of equipment and human ac-
In Fig. 1.8 the process control activities are organized tions. One layer includes process control functions, such
in the form of a hierarchy with required functions at as alarm management during abnormal situations, and
the lower levels and desirable but optional functions at safety instrumented systems for emergency shutdowns.
the higher levels. The time scale for each activity is The safety equipment (including sensors and control
shown on the left side. Note that the frequency of exe- valves) operates independently of the regular instru-
cution is much lower for the higher-level functions. mentation used for regulatory control in Level3a. Sensor
validation techniques can be employed to confirm that
Measurement and Actuation (Levell) the sensors are functioning properly.

Measurement devices (sensors and transmitters) and

actuation equipment (for example, control valves) Regulatory Control (Level 3a)
are used to measure process variables and implement As mentioned earlier, successful operation of a process
the calculated control actions. These devices are in- requires that key process variables such as flow rates,
terfaced to the control system, usually digital control temperatures, pressures, and compositions be operated
equipment such as a digital computer. Clearly, the at or close to their set points. This Level 3a activity,
regulatory control, is achieved by applying standard
feedback and feedforward control techniques (Chap-
ters 11-15). If the standard control techniques are not
(days-months) satisfactory, a variety of advanced control techniques
are available (Chapters 16-18). In recent years, there
has been increased interest in monitoring control sys-
tem performance (Chapter 21).
(hours-days)

Multivariable and Constraint Control (Level3b)

3b. Multivariable Many difficult process control problems have two dis-
(minutes-hours) and constraint tinguishing characteristics: (i) significant interactions
control
occur among key process variables, and (ii) inequality
constraints exist for manipulated and controlled vari-
ables. The inequality constraints include upper and
(seconds-minutes)
lower limits. For example, each manipulated flow rate
has an upper limit determined by the pump and control
valve characteristics. The lower limit may be zero, or a
2. Safety and small positive value, based on safety considerations.
environmental/
(< 1 second) Limits on controlled variables reflect equipment con-
equipment
protection straints (for example, metallurgical limits) and the oper-
ating objectives for the process. For example, a reactor
temperature may have an upper limit to avoid unde-
(< 1 second) sired side reactions or catalyst degradation, and a lower
limit to ensure that the reaction(s) proceed.
The ability to operate a process close to a limiting con-
straint is an important objective for advanced process
control. For many industrial processes, the optimum op-
erating condition occurs at a constraint limit-for exam-
Figure 1.8 Hierarchy of process control activities. ple, the maximum allowed impurity level in a product
 1.6 An Overview of Control System Design 9

Limit Limit Figure 1.9 Process variability

--Average, over time: (a) before improved
Controlled A2 process control; (b) after.
-·Average,
variable
Al

Time Time
(a) (b)

stream. For these situations, the set point should not be Also, the process model can be updated using parameter
the constraint value, because a process disturbance could estimation techniques and recent plant data (Chapter 7).
force the controlled variable beyond the limit. Thus, the
set point should be set conservatively, based on the abil- Planning and Scheduling (Level 5)
ity of the control system to reduce the effects of distur-
The highest level of the process control hierarchy is
bances. This situation is illustrated in Fig. 1.9. For (a),
concerned with planning and scheduling operations for
the variability of the controlled variable is quite high,
the entire plant. For continuous processes, the produc-
and consequently, the set point must be specified well
tion rates of all products and intermediates must be
below the limit. For (b), the improved control strategy
planned and coordinated, based on equipment con-
has reduced the variability; consequently, the set point
straints, storage capacity, sales projections, and the op-
can be moved closer to the limit, and the process can be
eration of other plants, sometimes on a global basis.
operated closer to the optimum operating condition.
For the intermittent operation of batch and semi-batch
The standard process control techniques of Level 3a
processes, the production control problem becomes a
may not be adequate for difficult control problems that
batch scheduling problem based on similar consider-
have serious process interactions and inequality con-
ations. Thus, planning and scheduling activities pose
straints. For these situations, the advanced control
difficult optimization problems that are based on both
techniques of Level 3b, multivariable control and con-
engineering considerations and business projections.
straint control, should be considered. In particular, the
model predictive control (MPC) strategy was developed
Summary of the Process Control Hierarchy
to deal with both process interactions and inequality
constraints. MPC is the subject of Chapter 20. The activities of Levels 1, 2, and 3a in Fig. 1.8, are re-
quired for all manufacturing plants, while the activities
in Levels 3b-5 are optional but can be very profitable.
Real-time Optimization (Level 4)
The decision to implement one or more of these
The optimum operating conditions for a plant are de- higher-level activities depends very much on the appli-
termined as part of the process design. But during plant cation and the company. The decision hinges strongly
operations, the optimum conditions can change fre- on economic considerations (for example, a cost/bene-
quently owing to changes in equipment availability, fit analysis), and company priorities for their limited re-
process disturbances, and economic conditions (for ex- sources, both human and financial. The immediacy of
ample, raw material costs and product prices). Conse- the activity decreases from Levell to LevelS in the hi-
quently, it can be very profitable to recalculate the erarchy. However, the amount of analysis and the com-
optimum operating conditions on a regular basis. This putational requirements increase from the lowest level
Level 4 activity, real-time optimization (RTO), is the to the highest level. The process control activities at
subject of Chapter 19. The new optimum conditions are different levels should be carefully coordinated and re-
then implemented as set points for controlled variables. quire information transfer from one level to the next.
The RTO calculations are based on a steady-state The successful implementation of these process control
model of the plant and economic data such as costs and activities is a critical factor in making plant operation
product values. A typical objective for the optimization as profitable as possible.
is to minimize operating cost or maximize the operat-
ing profit. The RTO calculations can be performed for
1.6 AN OVERVIEW OF CONTROL SYSTEM
a single process unit and/or on a plantwide basis.
DESIGN
The Level 4 activities also include data analysis to
ensure that the process model used in the RTO calcula- In this section, we introduce some important aspects of
tions is accurate for the current conditions. Thus, data control system design. However, it is appropriate first to
reconciliation techniques can be used to ensure that describe the relationship between process design and
steady-state mass and energy balances are satisfied. process control.
10 Chapter 1 Introduction to Process Control

Traditionally, process design and control system de- process. As mentioned earlier, process control should
sign have been separate engineering activities. Thus, in be based on process understanding.
the traditional approach, control system design is not ini- The major steps involved in designing and installing a
tiated until after plant design is well underway, and control system using the model-based approach are
when major pieces of equipment may even have been shown in the flow chart of Fig. 1.10. The first step, for-
ordered. This approach has serious limitations, because mulation of the control objectives, is a critical decision.
the plant design determines the process dynamics as well The formulation is based on the operating objectives for
as the operability of the plant. In extreme situations, the the plants and the process constraints. For example, in
process may be uncontrollable, even though the design the distillation column control problem, the objective
appears satisfactory from a steady-state point of view. A might be to regulate a key component in the distillate
more desirable approach is to consider process dynamics stream, the bottoms stream, or key components in both
and control issues early in the process design. The inter- streams. An alternative would be to minimize energy
action between process design and control is analyzed in consumption (e.g., heat input to the reboiler) while
more detail in Chapters 13, 25, and 26. meeting product quality specifications on one or both
Next, we consider two general approaches to control product streams. The inequality constraints should in-
system design: clude upper and lower limits on manipulated variables,
conditions that lead to flooding or weeping in the col-
1. Traditional Approach. The control strategy and
umn, and product impurity levels.
control system hardware are selected based on
After the control objectives have been formulated, a
knowledge of the process, experience, and insight.
dynamic model of the process is developed. The dynamic
After the control system is installed in the plant,
model can have a theoretical basis, for example, physical
the controller settings (such as controller gain Kc
and chemical principles such as conservation laws and
in Eq. 1-4) are adjusted. This activity is referred to
rates of reactions (Chapter 2), or the model can be de-
as controller tuning.
veloped empirically from experimental data (Chapter 7).
2. Model-Based Approach. A dynamic model of the
If experimental data are available, the dynamic model
process is first developed that can be helpful in at
should be validated, with the data and the model accu-
least three ways: (i) it can be used as the basis for
racy characterized. This latter information is useful for
model-based controller design methods (Chapters
control system design and tuning.
12 and 14), (ii) the dynamic model can be incorpo-
The next step in the control system design is to devise
rated directly in the control law (for example,
an appropriate control strategy that will meet the con-
model predictive control), and (iii) the model can
trol objectives while satisfying process constraints. As
be used in a computer simulation to evaluate alter-
indicated in Fig. 1.10, this design activity is both an art
native control strategies and to determine prelimi-
and a science. Process understanding and the experi-
nary values of the controller settings.
ence and preferences of the design team are key factors.
In this book, we advocate the philosophy that, for Computer simulation of the controlled process is used
complex processes, a dynamic model of the process to screen alternative control strategies and to provide
should be developed so that the control system can be preliminary estimates of appropriate controller settings.
properly designed. Of course, for many simple process Finally, the control system hardware and instrumen-
control problems, controller specification is relatively tation are selected, ordered, and installed in the plant.
straightforward and a detailed analysis or an explicit Then the control system is tuned in the plant using the
model is not required. For complex processes, how- preliminary estimates from the design step as a start-
ever, a process model is invaluable both for control sys- ing point. Controller tuning usually involves trial-and-
tem design and for an improved understanding of the error procedures as described in Chapter 12.

SUMMARY
In this chapter we have introduced the basic concepts Two physical examples, a continuous blending sys-
of process dynamics and process control. The process tem and a distillation column, have been used to intro-
dynamics determine how a process responds during duce basic control concepts, notably, feedback and
transient conditions, such as plant start-ups and shut- feedforward control. We also motivated the need for a
downs, grade changes, and unusual disturbances. systematic approach for the design of control systems
Process control enables the process to be maintained for complex processes. Control system development
at the desired operating conditions, safely and effi- consists of a number of separate activities that are
ciently, while satisfying environmental and product shown in Fig. 1.10. In this book we advocate the design
quality requirements. Without effective process con- philosophy that for complex processes, a dynamic
trol, it would be impossible to operate large-scale in- model of the process should be developed so that the
dustrial plants. control system can be properly designed.
 Exercises 11

Figure 1.10 Major steps in control

Information from
Formulate Management system development.
existing plants
control objectives objectives
(if available)

Q =Engineering activity

D =Information base

Physical
and chemical
principles
Plant data
(if available)
Process control
theory

Experience with
existing plants
(if available)

Vendor
information

Install control
system

Fl NAL CONTROL
SYSTEM

A hierarchy of process control activities was pre- Real-time optimization can be employed to determine
sented in Fig. 1.8. Process control plays a key role in the optimum controller set points for current operating
ensuring process safety and protecting personnel, conditions and constraints. The highest level of the
equipment, and the environment. Controlled variables process control hierarchy is concerned with planning
are maintained near their set points by the application and scheduling operations for the entire plant. The dif-
of regulatory control techniques and advanced control ferent levels of process control activity in the hierarchy
techniques such as multivariable and constraint control. are related and should be carefully coordinated.

EXERCISES
1.1 Which of the following statements are true? (c) Feedforward control can be perfect in the theoretical
(a) Feedback and feedforward control both require a mea- sense that the controller can take action via the manipulated
sured variable. variable even while the controlled variable remains equal to
(b) The process variable to be controlled is measured in its desired value.
feedback control.
12 Chapter 1 Introduction to Process Control

(d) Feedforward control can provide perfect control; that is, (b) Suggest a feedforward control method and sketch the
the output can be kept at its desired value, even with an im- schematic diagram.
perfect process model.
(e) Feedback control will always take action regardless of
the accuracy of any process model that was used to design it
and the source of a disturbance.
1.2 Consider a home heating system consisting of a natural c
gas-fired furnace and a thermostat. In this case the process 0
L
consists of the interior space to be heated. The thermostat F,z
u
contains both the measuring element and the controller. The M
furnace is either on (heating) or off. Draw a schematic diagram N
for this control system. On your diagram, identify the controlled
variables, manipulated variables, and disturbance variables. Be
sure to include several possible sources of disturbances that can
affect room temperature.
1.3 In addition to a thermostatically operated home heating
system, identify two other feedback control systems that can be
found in most residences. Describe briefly how each of them
works; include sensor, actuator, and controller information. 1.7 Two flow control loops are shown in the drawing. Indi-
cate whether each system is either a feedback or a feedfor-
1.4 Does a typical microwave oven utilize feedback control ward control system. Justify your answer. It can be assumed
to set cooking temperature or to determine if the food is that the distance between the flow transmitter (FT) and the
"cooked"? If not, what mechanism is used? Can you think of control valve is quite small in each system.
any disadvantages to this approach, for example, in thawing
and cooking foods?
1.5 Driving an automobile safely requires a large amount of
individual skill. Even if not generally recognized, the driver
needs an intuitive ability to utilize feedforward and feedback
control methods.
(a) In the process of steering a car, the objective is to keep
the vehicle generally centered in the proper traffic lane. Thus, System I
the controlled variable is some measure of that distance. If so,
how is feedback control used to accomplish this objective?
Identify the sensor(s), the actuator, how the appropriate con-
trol action is determined, and some likely disturbances.
(b) The process of braking/accelerating an auto is highly
complex, requiring the skillful use of both feedback and feed-
forward mechanisms to drive safely. For feedback control, the
System II
driver normally uses distance to the vehicle ahead as the mea-
sured variable. The "set point" then is often recommended to
be some distance related to speed, for example, one car length 1.8 I. M. Appelpolscher, supervisor of the process control
separation for each 10 mph. If this assertion is correct, how group of the Ideal Gas Company, has installed a 25 X 40 X
does feedforward control come into the accelerating/braking 5-ft swimming pool in his backyard. The pool contains
process when one is attempting to drive in traffic at a constant level and temperature sensors used with feedback con-
speed? In other words, what other information-in addition trollers to maintain the pool level and temperature at de-
to distance separating the two vehicles, which obviously should sired values. Appelpolscher is satisfied with the level
never equal zero-does the driver utilize to avoid colliding control system, but he feels that the addition of one or
with the car ahead? more feedforward controllers would help maintain the
pool temperature more nearly constant. As a new member
1.6 The distillation column shown in the drawing is used to of the process control group, you have been selected to
distill a binary mixture. Symbols x, y, and z denote mole frac- check Appelpolscher's mathematical analysis and to give
tions of the more volatile component, while B, D, R, and F your advice. The following information may or may not be
represent molar flow rates. It is desired to control distillate pertinent to your analysis:
composition y despite disturbances in feed flow rate F. All (i) Appelpolscher is particular about cleanliness and thus
flow rates can be measured and manipulated with the excep- has a high-capacity pump that continually recirculates the
tion of F, which can only be measured. A composition ana- water through an activated charcoal filter.
lyzer provides measurements of y. (ii) The pool is equipped with a natural gas-fired heater that
(a) Propose a feedback control method and sketch the adds heat to the pool at a rate Q(t) that is directly propor-
schematic diagram. tional to the output signal from the controller p(t).
 Exercises 13

(iii) There is a leak in the pool, which Appelpolscher has deter- (b) What additional variable(s) would have to be measured
mined is constant equal to F (volumetric flow rate). The liquid- to add feedforward control to the existing pool temperature
level control system adds water from the city supply system to feedback controller?
maintain the level in the pool exactly at the specified level. The (c) Write a steady-state energy balance. How can you de-
temperature of the water in the city system is Tw, a variable. termine which of the disturbance variables you listed in part
(iv) A significant amount of heat is lost by conduction to the (a) are most/least likely to be important?
surrounding ground, which has a constant, year-round tem- (d) What recommendations concerning the prospects of
perature T0 . Experimental tests by Appelpolscher showed adding feedforward control would you make to Appelpolscher?
that essentially all of the temperature drop between the pool
and the ground occurred across the homogeneous layer of 1.9 In a thermostat control system for a home heating system
gravel that surrounded his pool. The gravel thickness is Lh, (a) Identify the manipulated variable
and the overall thermal conductivity is k 0 . (b) Identify the controlled variable
(v) The main challenge to Appelpolscher's modeling ability (c) How is a valve involved in the control system? What
was the heat loss term accounting for convection, conduction, does it manipulate?
radiation, and evaporation to the atmosphere. He determined (d) Name one important disturbance (it must change with
that the heat losses per unit area of open water could be repre- respect to time).
sented by
1.10 Identify and describe three automatic control systems in
a modern automobile (besides cruise control).
1.11 In Figure 1.2 (h), identify the controlled, manipulated,
where
and disturbance variables (there may be more than one of
Tp = temperature of pool each type). How does the length of time for the dialysis treat-
Ta = temperature of the air, a variable ment affect the waste concentration?
U = overall heat transfer coefficient 1.12 For the steam-heated tank shown below, identify manip-
Appelpolscher's detailed model included radiation losses and ulated, controlled, and disturbance variables. What distur-
heat generation due to added chemicals, but he determined bances are measured for feedforward control? How would
that these terms were negligible. the control system react to an increase in feed temperature in
(a) Draw a schematic diagram for the pool and all control order to keep the tank temperature at its setpoint?
equipment. Show all inputs and outputs, including all distur-
bance variables.

,---------------
1
Feedforward Tsp
Figure E1.12. Feedforward control with a
feedback control loop for outlet temperature.
I r-------- controller (Entered manually)
1 I
I I
~I l~m I
I I I
I I I
I I I
I I IP

00I I
~------------
ri--~~------~~--~L~~------,
i

w I
I
I

-----~ ~~~.....f.-~~~1---~ T
w
I
I

Condensate
Steam
y
Chapter 2

Theoretical Models of
Che01ical Processes
CHAPTER CONTENTS
2.1 The Rationale for Dynamic Process Models
2.1.1 An Illustrative Example: A Blending Process
2.2 General Modeling Principles
2.2.1 Conservation Laws
2.2.2 The Blending Process Revisited
2.3 Degrees of Freedom Analysis
2.4 Dynamic Models of Representative Processes
2.4.1 Stirred-Tank Heating Process: Constant Holdup
2.4.2 Stirred-Tank Heating Process: Variable Holdup
2.4.3 Electrically Heated Stirred Tank
2.4.4 Steam-Heated Stirred Tank
2.4.5 Liquid Storage Systems
2.4.6 The Continuous Stirred-Tank Reactor (CSTR)
2.4.7 Staged Systems (a Three-Stage Absorber)
2.4.8 Distributed Parameter Systems (the Double-Pipe Heat Exchanger)
2.4.9 Fed-Batch Bioreactor
2.5 Solution of Dynamic Models and the Use of Digital Simulators
Summary

In this chapter we consider the derivation of unsteady- 2.1 THE RATIONALE FOR DYNAMIC
state models of chemical processes from physical and PROCESS MODELS
chemical principles. Unsteady-state models are also re- Dynamic models play a central role in the subject of
ferred to as dynamic models. We first consider the ratio- process dynamics and control. The models can be used to:
nale for dynamic models and then present a general
1. Improve understanding of the process. Dynamic
strategy for deriving them from first principles such as models and computer simulation allow transient
conservation laws. Then dynamic models are developed process behavior to be investigated without hav-
for several representative processes. Finally, we de- ing to disturb the process. Computer simulation
scribe how dynamic models that consist of sets of ordi- allows valuable information about dynamic and
nary differential equations and algebraic relations can steady-state process behavior to be acquired, even
be solved numerically using computer simulation. before the plant is constructed.

14