--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

3 -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

PROCESS CONTROL

O
n typical grass roots, chemical processing strategies, interlocks, sequencing and batch recipes, as well as
facilities, as much as 10% of the total capital display interfaces, alarm management, and Ethernet
investment is allocated to process control communication to networked computers, which are used to
equipment, design, implementation and provide supervisory control, inferential measures, data
commissioning. Process control is a very broad topic with historians, performance monitoring, and process
many distinct aspects. The following list of possible optimization.
sub-topics gives some idea of the full breadth of this topic: Also, the design practice includes P&ID documentation,
In the field, the topic includes the selection and database specification and verification of purchased
installation of sensors, transmitters, transducers, actuators, equipment, control design and performance analysis, software
valve positioners, valves, variable-speed drives, switches configuration, real-time simulation for DCS system checkout
and relays, as well as their air supply, wiring, power, and operator training, reliability studies, interlock classification
grounding, calibration, signal conditioning, bus architecture, and risk assessment of safety instrumented systems (SIS), and
communications protocol, area classification, intrinsic hazard and operability (HAZOP) studies.
safety, wired interlocks, maintenance, troubleshooting and Books have been written about each of these sub-topics
asset management. and many standards exist to specify best practices or provide
In the control room, the topic encompasses the selection guidance. The Instrumentation, Systems and Automation
and installation of panel mounted alarms, switches, recorders Society (ISA) is the primary professional society that addresses
and controllers, as well as Program Logic Controllers (PLC) and many of these different aspects of process control. The focus
Distributed Control Systems (DCS), including analog and of this chapter will be on control loop principles, loop tuning
digital input/output hardware, software to implement control and basic control strategies for continuous processes.

3.1. THE FEEDBACK CONTROL LOOP Control loop performance is determined by the response char-
acteristics of the block elements in the loop: the controller, valve,
Feedback control utilizes a loop structure with negative feedback to
process and measurement. Design choices can be made for the
bring a measurement to a desired value, or setpoint. A block diagram
valve, process and measurement, which can improve the achievable
of a typical process control loop is shown in Figure 3.1, with key
performance of the loop. The controller may then be tuned for the
elements of the loop being the controller, valve, process, and measure-
best performance of the resulting control loop, but must also
ment. Note that in addition to the setpoint entering the loop, there is
provide an operating margin from control instability. The control-
also a load shown. Changes in setpoint move the process to a new
ler tuning always establishes a trade off between resulting loop
value for the controlled variable, whereas changes in load affect the
performance and robustness due to this operating margin.
process resulting in a disturbance to the controlled variable.
The control loop must respond to either a change in setpoint or
OVERALL RESPONSE CHARACTERISTICS
a change in the load, by manipulating the valve in a manner that
affects the process and restores the controlled variable to its set There are both steady-state and dynamic response characteristics
point. Reacting to setpoint changes is called servo operation, and that affect loop performance. Steady-state gain is the most basic
reacting to load changes is called regulator operation. A flow con- and important of these response characteristics. Gain for a block
trol loop is a simple process example where both servo and regu- element can be simply defined as the ratio of change in output to a
lator operation is often required. The flow setpoint may be changed change in input. For several blocks in series, the resulting overall
to establish a new production rate. However, once set, it must be gain is the product of the individual block gains.
maintained during load changes, which disturb the flow through Dynamic responses can be divided into the categories of self-
the valve by altering upstream or downstream pressures. regulating and non self-regulating. A self-regulating response has

Load

Controlled
Setpoint Variable
1 Controller Valve 1 Process
2

Measurement

Figure 3.1. Block diagram of a control loop.

31
Copyright ß 2010 Elsevier Inc. All rights reserved.
DOI: 10.1016/B978-0-12-372506-6.00003-4
32 PROCESS CONTROL
inherent negative feedback and will always reach a new steady-state ler output to the process. Variable-speed drives provide fast and
in response to an input change. Self-regulating response dynamics linear response with little or no hysteresis and therefore are an
can be approximated with a combination of a deadtime and a first- excellent choice with respect to control performance. As the initial
order lag with an appropriate time constant. cost of variable-speed drives continues to decrease, their use should
Non self-regulating responses may be either integrating or run- become a more widespread practice.
away. An integrating response continues to change due to a lack of
inherent feedback. Since the output of an integrating response PROCESS CHARACTERISTICS
continues to change, its ‘‘steady-state’’ gain must be determined
as the ratio of rate of change of the output to a change in the input. An agitated tank is often used as an example of a first-order lag
Its response dynamics can be approximated with a combination of process. However, mixing in real tanks falls far short of the ideal
a deadtime, a first-order lag, and a ramp. Self-regulating responses well-mixed tank. Real tanks have composition responses that are a
with a very large time constant, or a very large gain, can also be combination of a first-order lag and deadtime. If the pumping rate
approximated as a pseudo-integrator during the first portion of of the agitator (Fa) is known, the deadtime (Td) of the real tank may
their response. be estimated by the following equation: Td = V/(F+Fa), where V is
A run-away response continues to change at an increasing rate the volume of the tank and F is the flow through it.
due to inherent positive feedback. The response is exponential and Process responses often consist of multiple lags in series. When
may be thought of as a first-order lag with a negative time constant. these lags are non-interacting, the resulting response is predomi-
Run-away response dynamics may be approximated with a combi- nantly deadtime, varying linearly with the number of lags in series.
nation of deadtime, a first-order lag and a second, longer lag with a However when these lags are interacting, such as the trays on a
negative time constant. distillation column, the resulting response remains predominantly a
first-order lag with a time constant proportional to the number of
VALVE CHARACTERISTICS lags squared.
Other process characteristics that affect control performance
Control valves have unique characteristics of their own which can are both steady-state and dynamic non-linear behavior. Steady-state
significantly affect the performance of a loop. The steady-state gain non-linear behavior refers to the steady-state gain varying, depen-
of the valve relates controller output to a process flow. How this dent upon operating point or time. For example, the pH of a process
flow affects the controlled variable of the process defines the range stream is highly non-linear, dependent upon the operating point on
of control. For servo control, the range of control would be defined the titration curve. Further, depending upon the stream component
as the range of setpoints achievable at a given load. For regulator composition, the titration curve itself may vary over time.
control, it would be defined as the range of loads for which the Non-linear dynamic behavior can occur due to operating point,
given setpoint could be maintained. Attempting to operate outside direction, or magnitude of process changes. For example, the time
the range of control will always result in the valve being either fully constant of the composition response for a tank will depend upon
open or closed and the controlled variable offset from setpoint. the operating point of liquid level in the tank. Some processes will
The steady-state gain of a control valve is determined at its respond in one direction faster than in the other direction, particu-
operating point, since its gain may vary somewhat throughout its larly as the control valve closes. For example, liquid in a tank may
stroke. Valves have internal trim that provide a specified gain as a drain quite rapidly, but once the drain valve closes the level can only
function of position, such as Linear, Equal Percentage, or Quick rise as fast as the inlet stream flow allows. The magnitude of a
Opening inherent characteristics. Typically, the trim is chosen such change may cause different dynamic response whenever inherent
that the installed characteristics provide an approximately linear response limits are reached. Process examples may include a transi-
flow response. Thus for a valve operating with critical gas flow, tion to critical flow, or a transition from a heat transfer to a mass
Linear trim would provide an approximately linear flow response. transfer limiting mechanism in a drying processes.
An Equal Percentage trim may be used to provide a more linear These non-linearities are the main reason an operating margin
response for gas or liquid flow where line pressure drop is equal or must be considered when tuning the controller. If the loop is to be
greater than the valve pressure drop. The Quick Opening trim is robust and operate in a stable manner over a wide range of condi-
usually not chosen for linear response in continuous control appli- tions, conservative values of the tuning parameters must be chosen.
cations, however, it provides a high gain near the closed position, Unfortunately, this results in poorer performance under most con-
which is useful for fast responding pressure relief applications. ditions. One technique to handle known non-linearities is to pro-
One common non-linear characteristic of control valves is vide tuning parameters that vary based upon measured process
hysteresis, which results in two possible flows at a given valve conditions.
position, depending upon whether the valve is opening or closing.
In the steady-state, hysteresis limits resolution in achieving a spe- MEASUREMENT CHARACTERISTICS
cific flow with its desired effect on the process. Dynamically, hys-
teresis also creates pre-stoke deadtime, which contributes to total Sensor type and location as well transmitter characteristics,
loop deadtime, thus degrading the performance of the loop. Pre- noise, and sampled data issues also can affect loop performance.
stroke deadtime is the time that elapses as the controller output Most continuous measurement sensors and transmitters have
slowly traverses across the dead band before achieving any change relatively fast dynamics and a noise filter, which can be approxi-
in actual valve position or flow. mated by a first-order lag with a one or two second time constant.
The use of a valve positioner can significantly reduce both Temperature sensors are somewhat slower as the sensor is in a
hysteresis and thus pre-stroke deadtime. A valve positioner is recom- thermowell, and these measurements have a larger, 15–30 second
mended for all control loops requiring good performance. Typical time constant.
hysteresis may be 2–5% for a valve without a positioner, 0.5–2% for Noise is often a problem in flow, pressure, and level measure-
a valve with an analog positioner, and 0.2–0.5% for a valve with a ments. Because flow is a very fast loop, controller tuning can be set
digital positioner. to ignore noise by using low gain and rely on a large amount of reset
On some control loops, a variable-speed drive on a pump, fan to take significant action only on sustained deviations. On slower,
or blower may be used as the final element connecting the control- non self-regulating loops like level, noise in the measurement can
 3.2. CONTROL LOOP PERFORMANCE AND TUNING PROCEDURES 33
degrade potential control performance by preventing the use of Another, perhaps the most important, controller parameter is
higher gains and/or derivative action in the controller. the control action, which is set as either ‘‘direct’’ or ‘‘reverse’’. If not
Excessive filtering of a signal to reduce noise would add effec- set correctly, positive feedback in the control loop would result in
tive deadtime to the loop, thus degrading the loop performance. unstable operation with the valve reaching a wide open or closed
One technique for reducing high amplitude, high frequency noise, limit. By convention, if the valve position is to increase as the
without introducing an excessive lag, is to rate limit the signal to a measurement increases, then the controller is considered ‘‘direct’’
rate comparable to the largest physically realizable upset. This acting.
approach chops off peak noise and allows a smaller time constant By first determining the process action, then specifying the
filter to effectively reduce the remaining lower amplitude, high opposite controller action, the desired negative feedback loop is
frequency noise. achieved. A typical flow loop is a good example as follows: the
Non-continuous measurements, such as produced by the process action is ‘‘direct’’ because the flow increases as the valve
sample and hold circuitry of a chromatograph, can introduce position is increased, therefore the controller action should be
significant deadtime into a loop. Also, the nature of the periodic specified as ‘‘reverse’’.
step change in value prevents the use of derivative action in the The actual output signal from the controller will further
controller. depend upon the specified failure mode of the valve. For example,
Distributed Control Systems often sample the transmitted a fail-closed valve will require an increase-to-open signal, whereas a
signal at a one second interval, sometimes faster or slower fail-open valve will require an increase-to-close signal. Most indus-
depending upon the characteristics of the process response. One trial controllers will have a separate parameter to specify the
concern related to sample data measurement is aliasing of the required signal for the failure mode of the valve. In order to mini-
signal, which can shift the observed frequency. However at a one mize confusion, rather than displaying actual output, most con-
second sample interval, this has seldom been a problem for all but trollers display an ‘‘implied valve position’’, which indicates the
the fastest process responses. A general rule for good perfor- desired position of the valve.
mance is to make the period between scans less than one-tenth The response characteristics of a direct acting PID controller
of the deadtime, or one-twentieth of the lag in the process are shown in Figure 3.2. For illustrative purpose, a step change to
response. the measurement is made and held constant without feedback. In
response to this disturbance, the independent contributions of each
CONTROLLER CHARACTERISTICS controller mode are provided in Figures 3.2(A, B and C), and the
combined PID response is presented in Figure 3.2(D). Note that the
The design of the valve, process, and measurement should be made Proportional mode has an immediate effect on the output, as
such as to minimize deadtime in the loop while providing a reliable, defined by its algebraic relationship. The Integral mode keeps
more linear response; then the controller can be tuned to provide changing the output at a constant rate as long as the constant
the best performance, with an acceptable operating margin for error persists. The Derivative mode provides an initial exaggerated
robustness. The PID controller is the most widespread and applic- response, which decays rapidly since the measurement stops chan-
able control algorithm, which can be tuned to provide near optimal ging after the initial step disturbance.
responses to load disturbances. PID is an acronym for Propor- Although there are many ways to implement PID modes into a
tional, Integral and Derivative modes of control. controller, the ISA standard algorithm is an ideal, non-interacting
Proportional mode establishes an algebraic relationship combination of the modes. This algorithm is a relatively new stan-
between input and output. The proportionality is set by a tunable dard, made feasible by digital implementation. Note that many
gain parameter. This unitless parameter, controller gain (Kc), speci- previously published tuning guidelines have been developed based
fies percent change in output divided by percent change in input. On upon various analog implementations of an interacting, series
earlier versions of PID controllers, an alternate parameter, Propor- combination of these modes.
tional Band (PB), was defined as the percent change in input
required to cause a 100 percent change in output. Thus by combining
definitions, these two terms are related as follows: Kc = 100/PB. 3.2. CONTROL LOOP PERFORMANCE AND TUNING
PROCEDURES
The Integral mode is sometimes referred to as ‘‘reset’’ because
it continues to take action over time until the error between mea- Any systematic tuning procedure must strive to provide optimal
surement and set point is eliminated. The parameter to specify this performance against some objective function. The first decision to
action is Integral time, which can be thought of as the length of time be made is whether this objective function is for setpoint response
for the controller to repeat the initial proportional response if the or load response. Optimizing setpoint response will result in slug-
error remained constant. Note that as this parameter is made gish load response, so if the primary objective of the loop is regula-
smaller, the reset increases as the control action is repeated in a tion, then the objective function should be a measure of load
shorter period of time. Some controllers use an alternate para- response performance.
meter, Reset, that is the reciprocal of Integral time and is referred A variety of criteria have been proposed for this objective
to as repeats/unit time. This latter approach is perhaps more intui- function such as the integral of square error (ISE), the integral of
tive in that as the Reset parameter is increased, there is more reset absolute error (IAE), or the integral of the time weighted absolute
action being applied. error (ITAE). The ISE criterion provides the greatest emphasis on
The Derivative mode is sometimes referred to as ‘‘rate’’ peak error, but is more oscillatory and less robust than the other
because it applies control action proportional to the rate of change criteria. Although for any given loop, ‘‘the beauty of the response is
of its input. Most controllers use the process measurement, rather in the eye of the beholder’’, in general the IAE criterion has become
than the error, for this input in order to not have an exaggerated the more widely accepted objective function to provide both
response to step changes in the set point. Also, noise in the process responsive and robust tuning.
measurement is attenuated by an inherent filter on the Derivative Numerous empirical correlations have been developed to
term, which has a time constant 1/8 to 1/10 of the Derivative time. determine PID tuning parameters for load responses of processes.
Even with these considerations, process noise is a major deterrent These correlations are based either on closed-loop procedures,
to the use of Derivative mode. which directly identify the ultimate gain and ultimate period of
34 PROCESS CONTROL
Implied Valve Position Implied Valve Position

Signal

Signal
Measurement Measurement

Setpoint Setpoint
Time Time
(A) Proportional Response (B) Integral Response

Implied Valve Position

Implied Valve Position

Signal

Signal
Measurement
Measurement

Setpoint
Setpoint
Time
Time
(C) Derivative Response
(D) PID Response

Figure 3.2. Response characteristics of a direct acting PID controller.

the loop, or on open loop procedures, which identify the time a pseudo time constant may be determined from the time it takes
constant and deadtime of a first-order plus deadtime approxima- the process variable, in percent, to move an amount equivalent to
tion of the process response. the percent change in controller output.
Correlations such as those presented in Table 3.2 may then be
CLOSED-LOOP PROCEDURE used to determine the values of tuning parameters based upon the
open-loop response (Edgar, 1999).
The closed-loop procedure requires tuning a controller with only
gain and increasing that parameter until sustained oscillations are DEFAULT TUNING
observed. The gain when this occurs is called the ultimate gain
(Ku) and the time between successive cycles is called the ultimate It is useful to have a set of robust, if not optimal, tuning parameters
period (Tu). for loops at startup. The values provided in Table 3.3 may be used
An alternative closed-loop approach called the ‘‘relay for that purpose. Loops with tuning outside the suggested range of
method’’ uses temporary narrow limits on the controller output values indicate either an unusual process or fundamental problems
and toggles between output limits each time the controller error with the valve, process, or measurement responses.
changes sign. The ultimate period is determined as before and the
ultimate gain is computed as Ku = 4*d/(3.14*a), where ‘‘d’’ is the 3.3. SINGLE STREAM CONTROL
range of the output limits, and ‘‘a’’ is the range of the process
measurement, both in percent. Flow, level, and pressure are process variables that can be con-
Correlations such as provided in Table 3.1 may then be used to trolled by manipulating their own process stream. Flow control is
determine the values of tuning parameters based upon the closed- typically used to establish throughput, whereas level and pressure
loop response (Edgar, 1999). are measures of liquid and gas inventory, which must be main-
tained to establish the overall process material balance. The process
OPEN-LOOP PROCEDURE material balance is typically controlled in the forward direction as
shown in Figure 3.3(A), where the feed flow rate to the process is
The open-loop procedure requires that the loop be placed in man- set, establishing the throughput and ultimate product rate after
ual mode and a step change in the controller output is made. The allowing for yield losses.
process response is recorded such that a time constant (Tc) and For the reaction area of a process, the large tank shown first in
deadtime (Td) may be determined from the data. The deadtime is these figures may be thought of as a shift or day tank, with its
the time before the process begins to respond. The time constant is inventory maintained by the periodic transfer of raw material into
the time it takes from the beginning of the process response until it it from outside the boundary limits of the process. For the refining
reaches approximately 63% of its final value. For non self-regulat- area of a process, it may be thought of as a large crude tank used to
ing processes, the deadtime is determined in the same manner, then isolate the crude and refining areas of the process. In either case, the

Table 3.1. Tuning Parameter Values from Closed-Loop Response

Controller Type Gain Integral Time Derivative Time

Proportional only, P 0.50*Ku – –

Proportional-Integral, PI 0.58*Ku 0.81*Tu –
Proportional-Integral-Derivative, PIDn 0.76*Ku 0.48*Tu 0.11*Tu
Proportional-Integral-Derivative, PIDi 0.55*Ku 0.39*Tu 0.14*Tu

Where: PIDn = non-interacting ISA algorithm; PIDi = interacting, series algorithm.

 3.3. SINGLE STREAM CONTROL 35
Table 3.2. Tuning Parameter Values from Open-Loop Response

Controller Type Gain Integral Time Derivative Time

Proportional only, P 0.56*Tc/Td – –

Proportional-Integral, PI 0.65*Tc/Td 3.5*Td –
Proportional-Integral-Derivative, PIDn 1.30*Tc/Td 2.1*Td 0.63*Td
Proportional-Integral-Derivative, PIDi 0.88*Tc/Td 1.8*Td 0.70*Td

Where: PIDn = non-interacting ISA algorithm; PIDi = interacting, series algorithm.

Table 3.3. Default and Range of Typical Tuning Parameter Values

Process Gain Integral Time (seconds) Derivative Time (seconds) Scan Period (seconds)

Liquid Flow/Pressure 0.3 (0.1–0.8) 6 (1–12) 0 (0–2) 1 (0.2–2)

Liquid Level 5.0 (0.5–20) 600 (120–6000) 0 (0–60) 2 (1–30)
Gas Pressure 5.0 (0.5–20) 300 (60–600) 0 (0–30) 1 (0.1–1)
Inline Blending 1.0 (0.1–10) 30 (10–60) 0 (0–30) 1 (0.5–2)
Exchanger Temperature 0.5 (0.1–10) 120 (30–300) 12 (6–120) 2 (0.5–5)
Column Temperature 0.5 (0.1–10) 300 (120–3000) 30 (6–600) 2 (1–30)
Reactor Temperature 2.0 (0.1–10) 600 (300–6000) 60 (6–600) 2 (1–10)
Inline pH 0.2 (0.1–0.3) 30 (12–60) 0 (0–6) 1 (0.2–2)
Neutralizer pH 0.2 (0.001–10) 300 (60–600) 60 (6–120) 2 (1–5)
Reactor pH 1.0 (0.001–50) 120 (60–600) 30 (6–60) 2 (1–5)

tank is sized large enough to provide continued operation of the their inlet flow and the units following would adjust their outlet
down stream equipment during short periods of interrupted supply. flow. Although these latter strategies are less common, they can
When such a tank is used as a transition between a batch and offer the advantage of fixing the feed to a specific unit that may
continuous process, it is desirable for the tank to hold at least otherwise be difficult to operate.
three batches of material.
By contrast, in Figure 3.3(B), a less common material balance
FLOW CONTROL
approach is taken, where the product flow rate is set directly and
each process unit must then adjust its inlet flow to maintain inven- Flow control is probably the most common control loop in most
tories. This approach is desirable when downstream factors fre- processes. Typically a liquid or gas flow rate is maintained in a pipe
quently determine the allowable production rate. This approach by a throttling valve downstream of the measurement as shown in
has the advantage that no yield assumptions are required in order Figure 3.4(A). Locating the valve upstream of the measurement is
to specify the production rate. not recommended because many measurement problems can arise.
Alternatively, an intermediate flow could be set as shown in Another method of controlling liquid flow is to adjust the
Figure 3.3(C), in which case the units ahead would have to adjust speed of a variable-speed drive on a pump as shown in Figure 3.4(B).
This approach is applicable to either centrifugal or positive displace-
Feed ment pumps and can provide significant energy savings at lower
rates because the power required is proportional to the speed
cubed. This approach also provides good control performance,
however a separate block valve is required to prevent leakage when
A B C the pump is stopped. Variable-speed drives have become much more
practical in recent years due to advanced electronics and micropro-
(A) Forward cessor developments, which allow variable frequency ‘‘vector’’ drives
for standard AC induction motors. In addition to providing precise
Product control and energy savings, these drives provide a soft start/stop and
do not require separate starting circuits.
Gas flow rate may also be controlled with variable-speed
drives on compressors, blowers or fans. The adjustment of louvers
A B C
or variable pitch fan blades, as shown in Figure 3.4(C), are addi-
tional methods for gas flow control. However, these latter devices
(B) Reverse have mechanical linkages that require high maintenance and intro-
duce significant hysteresis, which will degrade control loop perfor-
Intermediate
mance.
Solids may have their flow controlled by adjusting a motor speed
and inferring flow from the rate of displacement. Figure 3.4(D)
shows granular solids being flow controlled by a rotary vane
A B C feed valve at the bottom of a supply hopper. Figure 3.4(E) shows
the linear line speed of a belt feeder with a manually adjustable
(C) Mixed underflow weir height at the hopper. Figure 3.4(F) shows a rotary
feed plate, which controls solids flow by variable rotation
Figure 3.3. Material balance control. speed with a manually adjustable collar height and plow position.
36 PROCESS CONTROL
FC FT
FC FT
FT FC

SC

(A) Throttling Valve (B) Variable Speed Pump (C) Adjustable Louvers

(D) Rotary Valve (E) Belt Feeder (F) Rotary Plate

(G) Screw Feeder (H) Line Speed

Figure 3.4. Flow control.

Figure 3.4(G) shows a horizontal screw feeder or extruder, which acceptable. More typical is a setpoint either at the low end to
controls flow by adjusting the shaft speed. The flow of solids in the allow upstream equipment to keep running if the outlet flow
form of strings or sheets may be controlled by adjusting the line stops, or at the high end to provide feed for downstream equipment
speed of rollers as shown in Figure 3.4(H). if the inlet flow stops. For these latter cases, some integral action is
required to return the level to the setpoint. An error-squared PI
LEVEL CONTROL algorithm has proven effective for surge level control, providing
low gain near setpoint and proportionally higher gain at larger
Level control can be designed into the process with gravity, pres- deviations. In addition, logic that turns off the integral action
sure and elevation determining outlet flow. For example, the use of when the level is near setpoint can be helpful in eliminating slow
inlet and outlet weirs on the trays of a distillation column maintain continuous cycling.
both downcomer and tray levels as shown in Figure 3.5(A). For
operation at a pressure similar to downstream equipment, a sump PRESSURE CONTROL
level may be maintained by elevating external piping to provide a
seal, with a vent line to prevent siphoning as shown in Figure 3.5(B). Pressure in a pipe line may be controlled by manipulating either the
If downstream pressure is greater, then a barometric leg may be used inlet or outlet flow as shown in Figure 3.7(A). Pressure is an
to maintain a seal as shown in Figure 3.5(C). The overflow line integrating process, usually with negligible deadtime, therefore
shown must be adequately sized to self-vent, otherwise it may high gain and long integral time are recommended tuning. A pres-
begin to siphon. sure regulator is a self-contained valve and field controller with
For pumped systems such as shown in Figure 3.6, the tank high gain about a preset setpoint. Pressure regulators are often used
level may be controlled by manipulating either the outlet or inlet on plant utility streams such as instrument air or inert gas, the
flow. Direct control action is used when the outlet flow is adjusted. latter being shown to lower the pressure on the nitrogen supply in
Reverse action is required when the inlet flow is adjusted. Tank Figure 3.7(B).
level is an integrating process, usually with negligible deadtime, Pressure control of a tank at atmospheric conditions can be
therefore high gain and long integral time are recommended tuning achieved with a simple vent. However, often air cannot be allowed
when tight level control is desired. Tight level control is often to come into contact with the process, or volatile material cannot be
required for reactor and heat transfer vessels, but loose level control allowed to escape to the atmosphere. In these cases, an inert gas is
is preferred for surge tanks. used to ‘‘blanket’’ the material in the tank at a pressure slightly
The purpose of a surge tank is to reduce variations in the above atmospheric. Pressure control is achieved with split range
manipulated flow by absorbing the effect of temporary distur- control valves as shown in Figure 3.7(C). If liquid is withdrawn
bances. Ideally the tuning would be gain only, allowing the level from the tank, the pressure will decrease and the controller will
to vary about a mid-level setpoint with offset. However, in most open valve PV-1, allowing nitrogen to restore the pressure to set-
processes, a setpoint at mid-level and permanent offset are not point. If the tank fills with liquid, the pressure will increase and the
 3.4. UNIT OPERATION CONTROL 37

Outlet Weir

Vent line

Seal loop
Seal pan
Overflow line

(A) Weirs (B) Elevated Piping (C) Barometric Leg

Figure 3.5. Inherent level control.

Inlet Flow Vent

LT LC

Outlet Flow

(A) Forward

Inlet Flow Vent

LT LC

Outlet Flow

(B) Reverse

Figure 3.6. Level control.

controller will close valve PV-1, and then open PV-2 to let excess duced into the loop when the controller output must pass through
nitrogen out of the tank. the gap.
The graph in Figure 3.7(D) shows the relationship between
controller output and the valve positions. Sometimes a gap in the 3.4. UNIT OPERATION CONTROL
controller output about the point where both valves are closed will
be used to assure no overlap that would have both valve open at the Successful control loop implementation also requires a functional
same time. However, any gap should be minimized because the design strategy. A functional design strategy provides an equip-
pressure control performance will suffer as there is deadtime intro- ment layout and control loop interaction that best achieves the
38 PROCESS CONTROL
PT PC PC PT

(A) Forward (B) Reverse

PC

N2
PV-1 PV-2
PV-1 PV-2

Valve Position
PT

Controller Output
(C) “Inert Blanket” (D) Split Range Valves

Figure 3.7. Pressure control.

functional task required. The control strategy may apply a single thermal stresses would exist from the two different temperature
control loop or multiple control loops as required to achieve the streams. The three-way valve has flow characteristics similar to
functional objective of the unit operation. the fully overlapped two valves, which achieve an approximately
Multivariable model predictive control is often justified for constant resistance to total flow.
optimizing the performance of complex unit operations with sig- An interchanger would exchange heat between two process
nificant interactions and constraints. This type of control incorpo- streams, such as a pre-heater on a distillation column recovering
rates feedforward, decoupling, and constraint control into the heat from the bottom stream to the feed stream, or a pre-heater on a
design of the multi-loop controller. However, the performance boiler recovering heat from the stack gas to the combustion air. In
improvement achievable by that methodology remains highly these cases, the flow rates of the two process streams are set by
dependent on the proper design and implementation of the basic other control objectives and they are not available as manipulated
control system as is discussed in the following sections. variables. Only one process stream temperature can be controlled,
and this should be achieved with a bypass of that stream as pre-
HEAT EXCHANGERS WITHOUT PHASE CHANGE viously discussed.

Heat exchangers that exchange only sensible heat between the hot AIR COOLERS AND COOLING TOWERS
and cold streams may have one process stream and a utility stream
such as hot oil, cooling tower water, chilled water or air. The flow Air coolers and cooling towers often use multiple two-speed fans
rate of the utility stream is often adjusted to control the outlet and discrete control logic that steps the fan speeds progressively to
temperature of the process stream as shown in Figure 3.8(A). The adjust air flow in order to maintain a stream temperature. For
temperature response will be non-linear with deadtime and multiple example, consider a cooling tower as shown in Figure 3.9(A)
lags. The control performance will benefit from tuning with Deri- using four, 50-Hp fans, each capable of being set to off, half
vative action. The valve on the utility stream may be either on the speed, or full speed operation. Air flow is proportional to the fan
inlet or the return. Cooling tower water is best throttled at the inlet speed, while power consumed is proportional to speed cubed. There
as the cooler water is less likely to cavitate in the valve. Cooling would be 9 distinct air flows available for cooling, with a resolution
tower water should not be throttled to the extent that its return of 12.5%, providing rather coarse temperature control. With PID
temperature exceeds 120 degrees F, at which point fouling becomes control, the fan speeds will cycle continually as the temperature
a problem. oscillates above and below setpoint. Tight tuning will cycle the fan
Figure 3.8(B) shows an alternative control scheme including a speeds more frequently. Cycling a fan speed more than 4 times an
bypass of the process stream around the exchanger. This arrange- hour may be considered severe service for its motor, likely to incur
ment can offer much better temperature control, as now the tem- greater maintenance costs.
perature response is linear and fast, because the dynamics of the An equivalent area, cooling tower could be designed using two,
exchanger are no longer within the control loop. Note that the 100 Hp fans with variable-speed drives as shown in Figure 3.9(B).
process now being controlled is simply the linear, thermal blending Turndown operation would first decrease both fan speeds down to
of a hot and cold stream. their minimum speed, at approximately 12% of full speed. Then one
The two valves being adjusted by the temperature controller fan would shut off as the other fan doubled its speed, in order to
are implemented such that their stroking fully overlaps, with one maintain the air flow. The running fan would then decrease again to
valve closing as the other valve opens, as shown in Figure 3.8(C). its minimum speed before being shut off. On increasing operation,
With smaller pipe sizes, these two valves could economically be first one fan would start at 18% output and increase up to 36%, at
replaced with a three-way valve located at the start of the bypass which point the second fan would start and the controller output
line. It should not be located at the end of the bypass line where would reduce back to 18% for both fans to maintain air flow. Both
 3.4. UNIT OPERATION CONTROL 39
CTR

Process In

Process Out

TC TT

CTS
(A) Utility Flow

Process In CTR

TV-1

TC
TV-1 TV-2

Valve Position
TT

Process Out

TV-2 CTS Controller Output

(B) Process Bypass (C) Split Range Valves

Figure 3.8. Temperature control of heat exchangers without phase change.

Fan-1 Fan-2 Fan-3 Fan-4 %Air

Off Off Off Off 0
Lo Off Off Off 12.5
Lo Lo Off Off 25.0
Lo Lo Lo Off 37.5
Lo Lo Lo Lo 50.0
Hi Lo Lo Lo 62.5
Hi Hi Lo Lo 75.0
Hi Hi Hi Lo 87.5
Hi Hi Hi Hi 100

(A) Multiple 2-Speed Fans

SC SC

CTR

TC

TT
LC LT CTS

Make-up
(B) Variable Speed Fans

Figure 3.9. Temperature control of air cooled exchangers and cooling towers.
40 PROCESS CONTROL
fans would then be increased up to full speed if required. Note there refrigerant side is determined by the temperature that provides heat
is a gap between shutting off a fan at 12% and starting it back up at transfer equal to the heat absorbed by the boiling refrigerant. A
18%, which prevents cycling. The cost of the variable-speed drives level controller is shown in Figure 3.10(B) maintaining the liquid
would be offset by the simpler construction and no need for two level above the tubes of the exchanger.
speed fans and their start/stop circuitry. In addition, the variable- Measurement of the steam or refrigerant flow can provide a
speed drives would provide improved temperature control and less good indication of heat duty. If there are multiple users, which cause
power usage. For example, at 75% air flow, the two variable-speed disturbances to the utility, then a temperature to flow cascade con-
drives would use (100 + 100)* (.75)^3 = 84.4 Hp, whereas the 4 two- trol arrangement should be considered. In such a cascade arrange-
speed motors with two at High speed and two at Low speed would ment, the temperature controller output provides the setpoint for the
use (50 + 50)* (1.0)^3 + (50 + 50)*(.50)^3 = 112.5 Hp. flow controller. The flow controller minimizes the effect of utility
Make-up water must be added because the cooling tower has stream disturbances and linearizes the temperature control loop.
direct contact, evaporative cooling between the water and the air,
as well as losses due to droplets entrained into the air. Although not PROCESS CONDENSERS
shown, a continuous purge of water is required, because otherwise
impurities will build up as water evaporates. Also not shown are Condensing process vapor usually requires adjustment of the heat
chemical additives, which often are added periodically to treat the removal such that the amount of vapor condensed matches the vapor
water in order to retard bacterial and fungal activity. A level con- supply, in other words, pressure control. The most effective manner
trol loop is shown which adjusts the make-up water flow to main- of adjusting the heat removal is to vary the area available for
tain level as required by the resulting water balance. condensing. This may be accomplished by blocking a portion of the
total area with condensate or non-condensable gas (inerts). In
HEAT EXCHANGERS WITH PHASE CHANGE Figure 3.11(A), a condenser is shown with inerts blanking the lower
portion of its tubes. Vapor, as it is being condensed, pushes the inerts
A steam heater, as shown in Figure 3.10(A), can provide responsive to the far end of the condenser. When pressure rises due to more
temperature control because the entire steam side is at the conden- vapor arriving at the condenser, the inerts are pushed out, exposing
sing temperature and has a high heat transfer coefficient. The more tube area for condensing. When pressure drops due to less
pressure on the steam side is determined by the temperature that vapor arriving, inerts flow back into the condenser, blanking off
provides heat transfer equal to the heat released by the condensing more tube area.
steam. The steam trap provides condensate level control within the For a condenser operating at atmospheric pressure, an ade-
trap in order to provide a seal for the condensing steam. quate vent is all that is necessary. However, air is often not suitable
A refrigerant cooler likewise provides responsive temperature for contact with the process due to concern about contamination or
control because the entire refrigerant side is at the boiling tempera- flammability. In these cases, the vent may be connected to a source
ture and has a high heat transfer coefficient. The pressure on the of low pressure nitrogen, or other inert gas. For vacuum operation,
the vent must also be connected to a vacuum pump or steam jet
(eductor) as shown in Figure 3.11(B). The pressure controller adjusts
the split range control valves such that as its output decreases, first
PV-2 closes then PV-1 opens. Normal operation would have PV-1
Process In Steam closed and PV-2 open, therefore the inert gas is used only sparingly.
Coolant flow is generally not throttled for pressure control,
however it is occasionally adjusted for temperature control of the
TC
sub-cooled condensate. Unless there is significant sub-cooling, this
latter temperature loop is often ineffectual. At best, it will require
loose tuning, or often it will be placed into manual for seasonal
TT
adjustment only.
As mentioned previously, condensate may also be used to blank
Steam Trap T
off tube area for pressure control. The two methods shown in Figure
Process Out
3.12 may be used when inerts are not present in significant amounts.
(A) Steam Heater The first method shown in Figure 3.12(A) places a valve in
the condensate line and directly backs the liquid up into the con-
denser as needed. The second method places a control valve in a
vapor line, added to bypass the condenser and go directly into the
Process In Return accumulation tank (see Figure 3.12(B)). Although somewhat coun-
ter-intuitive, the vapor bypass valve must open as pressure drops,
in order to raise the pressure at the accumulation tank and force
TC
liquid condensate back into the condenser to restore the energy
balance. Care must be taken that the surface of the liquid in the
tank is not disturbed, as undesirable pressure transients can
develop with the hot vapor in contact with the sub-cooled liquid.
TT
Another concern with both of these condensate methods is the
LC LT
venting of process inerts as they build up over time. Valve position
Process Out is the best indication of this pending problem.

Refrigerant
PROCESS VAPORIZERS
(B) Refrigerant Cooler
A vaporizer is typically used in a process to provide a vapor feed
Figure 3.10. Temperature control of exchangers with phase change. to downstream equipment. In that case, it may be desirable to set
 3.4. UNIT OPERATION CONTROL 41

Tp
Tc

Qcond 5 UA(Tp-Tc)

CTS

Vent

Condensate
(A) Condenser Detail

PC
Motive Steam

N2

PV-1 PV-2
PT Steam Jet
Vapor Vent
CTR

Condensate
CTS
(B) Condenser with Steam Jet

Figure 3.11. Condenser pressure control with inert gas.

the flow of vapor directly with the setpoint of a flow control loop. rate. This design requires excess blade heater area, which also serves
Then the heat input is adjusted for pressure control and the liquid to provide superheat to the vapor.
level is maintained by adjusting the inlet flow, in a reverse material
balance manner as shown in Figure 3.13(A). If heat transfer limits EVAPORATORS
throughput, then both the vapor valve and the steam valve will
operate fully open and the pressure will droop to an equilibrium An evaporator provides one stage of separation based upon relative
point, where heat transfer equals the flow through the downstream volatility. It is typically used with systems having a large relative
equipment. volatility, such as salts and solvents. When water is being removed
For a direct material balance, the inlet liquid is flow controlled, as an overhead vapor, multi-effect operation often may be used to
the level is maintained by adjusting the steam flow and the pressure provide improved energy efficiency. Figure 3.14 shows two alter-
is then controlled by adjusting the vapor flow, as shown in native direct material balance evaporator control schemes.
Figure 3.13(B). If the feed pump limits capacity, the pressure should In both of these schemes, the feed is flow controlled and the
be allowed to droop by putting the controller into manual with a overhead vapor flow is adjusted for pressure control. The arrange-
fully open valve. Then the feed flow controller will run with its valve ment shown in Figure 3.14(A) is more common, with the level
fully open and the pressure will droop to an equilibrium point, where controller adjusting the bottom flow and the temperature control-
inlet flow equals the flow through the downstream equipment. ler adjusting the steam flow. Note that the temperature is an
Because of impurities in the liquid feed, it may be necessary to inferred measure of composition. This inferred composition con-
purge or blowdown the vaporizer periodically, as indicated by trol is achieved by adjusting the steam flow such that the material
a rising boiling point temperature or the steam valve approaching balance has more or less vapor removed overhead.
a full open position. Although not shown, vaporizers often have a If the bottom flow is a small fraction of the feed, then it will not
separate, temperature controlled superheater to ensure the vapor- provide satisfactory level control. A better arrangement for that
ization of any entrained droplets and prevent condensation. situation is the alternative shown in Figure 3.14(B), with the level
Smaller, low cost vendor packaged vaporizers often employ controller adjusting the steam flow and temperature controller
self-regulation of the level. These units typically have blade heaters, adjusting the bottom flow. As before, the temperature is an inferred
which allow the level to vary until the area available for heat measure of composition, which is controlled by adjusting the mate-
transfer to the liquid provides vaporization to match the feed rial balance split. When the temperature is above setpoint, implying
42 PROCESS CONTROL
PT PC

CTR
Vapor

TC

TT
CTS Condensate

(A) Valve in Condensate Line

PT PC
Vapor

Accumulation LT
Tank
CTR
LC

Condensate
CTS

(B) Valve in Vapor Bypass

Figure 3.12. Condenser pressure control with condensate.

that the high boiler composition is too high, the bottom flow must Another advantage of this scheme is that it lends itself readily
be increased to remove more of the high boiler. to the application of feedforward control in order to maintain the
D/F ratio for measured changes in feed flow. This feedforward
signal would be trimmed by the addition of a feedback signal
DISTILLATION COLUMNS
from the column temperature controller.
In simple distillation, a feed is separated into a distillate and bot- Scheme 2 indirectly adjusts the material balance through the
toms product with multiple stages of separation based upon relative two level control loops. If the reflux flow is increased, then the reflux
volatility. Both distillate and bottom composition may be con- accumulator level controller decreases the distillate flow. As the
trolled by adjusting the material balance split and the separation. additional liquid proceeds to flow down to the sump, the sump
However, basic column control schemes attempt to control only level controller increases the bottoms flow a like amount. The
one composition within the column by adjusting the material bal- separation is held constant by manually setting the reboiler steam
ance and simply fix the separation at an optimal value. flow to maintain a constant energy per unit feed. This scheme is
There are four alternative control schemes that are commonly recommended for columns with a small reflux ratio (R/D < 1). This
used for distillation column control as shown in Figure 3.15 scheme also offers improved dynamics, which may be required,
through Figure 3.18, respectively. Scheme 1 directly adjusts the particularly if the column has a large horizontal reflux accumulator.
material balance by manipulation of the distillate flow. If the dis- Scheme 3 indirectly adjusts the material balance through the
tillate flow is increased, then the reflux accumulator level controller two level loops. If the steam flow is increased, then the sump level
decreases the reflux flow. As less liquid proceeds to flow down to controller decreases the bottom flow. As the additional vapors go
the sump, the sump level controller decreases the bottoms flow a overhead and condense, the reflux accumulator level control
like amount. The separation is held constant by manually setting increases the distillate flow a like amount. The separation is held
the reboiler steam flow to maintain a constant energy per unit feed. constant by manually setting the reflux flow to maintain a relatively
This scheme is recommended when the distillate flow is one of constant energy per unit feed. This scheme is recommended for
the smaller flows in the column, particularly when the reflux ratio is columns with a small energy per unit feed (V/F < 2). This scheme
large (R/D > 3). Also, it is important that the reflux accumulator also offers the fastest dynamics.
level control can be tightly tuned and that the liquid holdup is not Scheme 4 directly adjusts the material balance by manipulation
too large (< 5 minutes). This scheme has the least interaction with of the bottom flow. If the bottom flow is decreased, then the sump
the energy balance, as it provides a good range of control with only level controller increases the reboiler steam flow. As the additional
small changes in the distillate flow, and it also provides a form of vapors go overhead and condense, the reflux accumulator level con-
automatic internal reflux control. If the reflux becomes more sub- trol increases the distillate flow a like amount. The separation is held
cooled, initially additional vapors will be condensed inside the constant by manually setting the reflux flow to maintain a relatively
column. However, the overhead vapors will be reduced by exactly constant energy per unit feed. This scheme is recommended when the
the same amount, and with tight level control, the reflux will then bottom flow is one of the smaller flows in the column, particularly
be reduced accordingly. when the bottom flow is less than 20% of the vapor boilup.
 3.4. UNIT OPERATION CONTROL 43

PC
PC

PT FT

Vapor

LC

LT

Steam
Liquid Feed

(A) Reverse Material Balance

PC

PT

Vapor

LC

FC

LT
FT

Steam
Liquid Feed

(B) Direct Material Balance

Figure 3.13. Vaporizer control.

Scheme 4 has little interaction with the energy balance, as it installed one stage above and below this location because of uncer-
provides a good range of control with only small changes in the tainty in tray efficiencies and VLE data.
bottom flow. However, the tuning of the sump level loop usually Product compositions from these same parametric cases can be
makes this scheme slower than the others. An inverse response is plotted against the temperature occurring on the uncontrolled
also possible with this sump level control loop. This type of control stage as shown in Figure 3.20. The steady-state effect on
response occurs when an increase in steam flow temporarily causes product composition due to temperature measurement errors can
the sump level to increase before it begins to decrease. If this occurs, be determined in this manner.
the level loop must be detuned even more.
An advantage of this scheme is that it lends itself readily to the LIQUID-LIQUID EXTRACTION
application of feedforward control in order to maintain the B/F
ratio for measured changes in feed flow. This feedforward signal Liquid-liquid extraction involves contacting of two immiscible
would be trimmed by the feedback signal from the temperature liquids and then subsequent separation by settling. Liquid-liquid
controller. extraction can take place in a column with various internals to
Once a basic column control scheme is chosen, simulation of foster contact between the dispersed and continuous liquid phases.
parametric steady-state cases can be used to determine the best Internals can include sieve trays, baffle trays and packing, as well as
temperature control stage location. These cases should hold the mechanical agitation and pulsing of the liquid.
separation variable constant and adjust the material balance in the The solvent flow is often maintained in proportion to the feed
manner of the chosen control scheme. Temperature profiles from flow. When the dispersed phase is the heavy phase, an interface will
these parametric cases can then be plotted together as shown in form towards the bottom of the column and is controlled as shown
Figure 3.19. The best control stage location is where the largest, in Figure 3.21. When the dispersed phase is the light phase, an
most symmetrical temperature deviation from the base case occurs. interface will then form towards the top of the column, however
On new columns, it is recommended that additional thermowells be the bottom flow is still manipulated to maintain the interface level.
44 PROCESS CONTROL

PC PC

PT PT
Vapor
TC TT
Vapor
TC
LC

TC
Steam Steam
LT LC LT

FC FC

T T
FT FT
Liquid
Liquid

Liquid Feed Liquid Feed

(A) Level Control by Adjusting Liquid (B) Level Control by Adjusting Steam

Figure 3.14. Evaporator control.

FC

FT
LC LT

Reflux

FC

FT TT TC FC

Feed FT

Distillate
FC

FT

Steam LT LC

FC

FT

T
Bottoms

Figure 3.15. Distillation column control—scheme 1.

One or more mixer-settlers in series can also be used to per- is not available for feedback correction of flow, therefore precise
form liquid-liquid extraction. Often a series of mixer settlers will be flow measurements are required. Mass flow corriolis meters are
installed on a gradient to allow gravity flows between vessels as ideal for feed line sizes below 6 inches. Multiple feeds should be
shown in Figure 3.22. Adjustable piping and sight glasses allow the ratioed off of one primary feed flow.
system to be set up to be self-regulating over a narrow range of feed A jacketed vessel as shown in Figure 3.23 is often used for
rates. Often the solvent is added in ratio to the feed. Different ratios maintaining temperature; however it has limited surface area and
and sometimes different solvents are used in subsequent stages. low heat transfer coefficients. Sometimes internal reactor cooling
coils are also used to provide additional heat transfer area. In order
REACTORS to manipulate the heat transfer, maximum flow is maintained in a
Reactor control is largely about maintaining stoichiometric ratios circulation loop, while the jacket temperature is adjusted by bring-
of feeds and temperature control. Often composition measurement ing in and letting out coolant.
 3.4. UNIT OPERATION CONTROL 45

FC

FT
LT LC

Reflux

FC

FT TT TC FC

FT
Feed

Distillate
FC

FT

Steam LT LC

FC

FT

T
Bottoms

Figure 3.16. Distillation column control—scheme 2.

FC

FT
LT LC

Reflux

FC

FT TT TC FC

Feed FT

Distillate
FC

FT

Steam LT LC

FC

FT

T
Bottoms

Figure 3.17. Distillation column control—scheme 3.

46 PROCESS CONTROL

FC

FT
LT LC

FC

FT TT TC FC

Feed FT

Distillate
FC

Inverse level response

FT

Steam LT LC

FC

FT

T
Bottoms time

Figure 3.18. Distillation column control—scheme 4.

Benzene-Toluene Column Temperature Profiles 100

18
99.8
17
16 99.6
DISTILLATE PRODUCT
15
99.4
Composition, wt %

14
99.2
13
12 99
11
Stage Number

98.8
10
9 +1%D/F 98.6
BOTTOM PRODUCT
8 98.4
7
98.2
6
5 98
85 90 95 100 105 110
4 21%D/F
Control Stage Temperature, deg F
3
2 Figure 3.20. Distillation column control—temperature sensitivity.
1
0
70.0 75.0 80.0 85.0 90.0 95.0 100.0 105.0 110.0 115.0 120.0
Temperature, Deg C side, while the heat transfer medium is throttled by the reactor
temperature controller. If the reactor is small and well mixed, the
Figure 3.19. Distillation column control—control plate location. cascade temperature control arrangement as shown may not be
necessary, and the reactor temperature controller may be con-
nected directly to the valve.
The reactor temperature controller provides a setpoint to the When the reaction temperature is high enough to vaporize the
jacket temperature controller. Heat transfer is linear and propor- reactants, an external condenser is an effective way to remove heat
tional to the temperature difference. as shown in Figure 3.25. The reactor pressure is adjusted to main-
Another approach for removing heat is a circulation loop tain the corresponding boiling temperature.
through an external heat exchanger as shown in Figure 3.24. The For fast reaction kinetics, the feed flow may also be adjusted to
circulation rate is maximized for good heat transfer on the process maintain temperature. In this case the rate of heat removal sets the
 3.4. UNIT OPERATION CONTROL 47
PC

PT

S/F
FY FC Raffinate

FT

Feed
FC

FT

Solvent
LT LC

Extract

Figure 3.21. Liquid-liquid extraction tower.

FC FY

FT
Solvent
FC FY

FT
FC
Solvent
FC FY
FT
FT
Solvent
Feed Raffinate

Raffinate

Raffinate

Extract

Extract

Extract

Figure 3.22. Liquid-liquid extraction with mixer/settlers.

production rate. For the fluid bed reactor shown in Figure 3.26, control strategy shown in Figure 3.27. The firing demand is often
a series of separate cooling coils may be manually put into service set by a steam pressure controller for boiler operation, or by a
to establish incremental production rate. process temperature controller for a fired heater. The firing
demand signal goes to a high selector (HS) for the air flow and to
COMBUSTION a low selector (LS) for the fuel flow.
The other signal going to the low selector is based upon the air
Combustion is the oxidation reaction of fuel with air, as occurs in flow converted to Btu/hr and increased by a slight offset. This offset
boilers and fired heaters. Maintaining the appropriate ratio of is tuned in order to not respond continually to noise in the air flow
oxygen to fuel is critical to the efficiency and safety of the flame measurement. At steady state, the low selector will select the firing
chamber. Too much oxygen wastes energy heating excess air, demand signal because of this offset. However, during a transient
whereas too little can result in an explosion. increase in firing demand, the signal based upon the air flow will be
In order to operate with a minimum of excess air, a cross selected to set the fuel flow, until the air flow increases to satisfy the
limiting scheme is often utilized to assure that during transients, increased demand.
the air flow will increase before the fuel flow, and the fuel flow will The other signal going to the high selector is based upon the
decrease before the air flow. This cross limiting is achieved by the fuel flow converted to Btu/hr and decreased by a slight offset tuned
48 PROCESS CONTROL
FC FY
B/A
FC TC

FT
FT

Feed B
PC
Feed
FC
PT
FT
PC
Feed A
Vent
PT
TC
TT
Vent
TT
LT LC

CTR

TC TT
LT LC

CTS Product

Figure 3.25. External condenser reactor control.

Product
only a "10% correction. In prior analog based control systems, the
Figure 3.23. Jacketed reactor control.
O2 controller would adjust the measured air flow signal, thus
maintaining a fixed factor with a pseudo flow. However, with
current digital based systems, the Btu/lb factors and their inverse
TC
PC may be computed accurately; therefore the measured air flow need
not be altered. The O2 controller setpoint must be chosen carefully
PT to provide an adequate excess to account for any incomplete mixing
TC TT of fuel and air at the burner. The excess required is dependent upon
Vent
the type of fuel and burner design.

CTR pH
TT pH measurement is difficult and often unreliable. When a high degree
LT LC of reliability is required, a three-probe system with an automatic mid
CTS selector is recommended as shown in Figure 3.28(C). Most titration
curves are highly non-linear with respect to pH but can be linearized
somewhat by converting the signal to reagent demand.
If the reagents react rapidly, then in-line control of pH is
FC practical, with a signal filter to reduce the noise. A pump or in-
Product line static mixer should be used to provide thorough mixing of the
FT process and reagent as shown in Figure 3.28(A) and (B).
Often existing agitated tanks merely provide enough circula-
tion to keep solids from settling; however, for pH control the
Feed
agitator must circulate liquid at a rate of 20 times the throughput
Figure 3.24. External heat exchange reactor control. to be considered well mixed. For tanks which are not well mixed, it
is better to provide in-line pH on the feed to the tank.

to reduce noise. At steady state the high selector will select the firing TURBINES AND COMPRESSORS
demand signal because of this offset. However, during a transient
decrease in firing demand, the signal based upon the fuel flow will Steam turbines are most often used in processes to provide power to
be selected to set the air flow, until the fuel flow decreases to satisfy compressors or electric generators as shown in Figure 3.29. Multi-
the reduced demand. stage turbines may also admit or extract steam between stages.
This scheme can be easily expanded to handle multiple fuels, in Turbine speed is a fast loop, controlled by manipulating the supply
which case the signal from the low selector, as total Btu/hr, would steam valve or valves, as often there will be a rack of parallel steam
be proportioned to the setpoints of additional fuel flow controllers, valves supplied as part of the turbine system.
each converted with their own Btu/lb fuel factor. Then the multiple Steam turbines with electrical generators are used to recover
fuel flow measurements would each be converted to Btu/hr and the power from high pressure steam when the plant steam balance
added together with the negative offset as the other input to the requires additional low pressure steam. As such, they provide more
high selector. energy efficient pressure regulation than a simple pressure control
An O2 controller can be used to adjust the Btu/lb Air factor valve. Admission or extraction of steam to a header between stages
(K), decreasing it to require more air if the measured oxygen is and/or condensation of exhaust steam fulfill additional steam bal-
below setpoint. Typically, the O2 controller output is limited to ance and energy recovery needs.
 3.4. UNIT OPERATION CONTROL 49
PC
PC

PT
PT

Product Steam

TC TT

LT LC

B/A
FY FC Feed Water

FT FT

FC
FC Feed A

FT
Blowdown

Feed B

Figure 3.26. Feed rate reactor control.

2
HS + Btu/hr Offset

Firing Demand 1/K

as Btu/hr

RSP
FC

Air
FT

K 5 Btu/lb Air, which can be

K adjusted by the output of an
O2 controller

LS 1 Btu/hr Offset

1/A A
A5 Btu/lb Fuel

RSP
FC

Fuel
FT

Figure 3.27. Cross limiting combustion control.

For electrical generators, at startup, the turbine speed must be adjusts the setpoint on the turbine speed controller. In smaller
ramped up until the generator is providing electrical cycles that are processes, fixed speed compressors may be used by adjusting either
synchronized with the power grid. Once connected to the grid, the an inlet or discharge valve to achieve pressure control. It is more
speed becomes essentially self regulating, and the electrical power energy efficient to adjust an inlet valve, or better yet to adjust inlet
generated varies directly with steam supplied. Steam header pres- vanes which provide a pre-rotation to the gas. However, adjust-
sures may then be controlled by adjusting the inlet steam, extrac- ment of speed is the most energy efficient method control.
tion/admittance steam and exhaust condenser. Both axial and centrifugal compressors are subject to an
Steam turbines with compressors are used for providing pro- unstable region of low flow operation called surge. This region is
cess gas flow at a required pressure in high throughput processes. defined by the pressure-flow relationship which has a peak deter-
The process demand is determined by a pressure controller, which mined by the operating speed. For various compressor speeds,
50 PROCESS CONTROL
Reagent Reagent

PC pHT PC pHT

Feed Feed

(A) Static Mixing (B) Pump Mixing

pHC

Feed Reagent

FY

pHT

pHT

pHT

(C) Agitated Tank

Figure 3.28. pH Control.

FC FY

FT Outlet
PT

Inlet

SC ST

PC

Supply Steam

Exhaust Steam

Figure 3.29. Compressor/Turbine Control.

 REFERENCES 51
300.0
Surge Line Inputs from the pressure and speed transmitters are used to com-
100 %Speed pute the required minimum flow setpoint, based upon an anti-surge
control line relationship shown in Figure 3.30. The anti-surge con-
250.0 trol line is determined by applying an operating margin to the
90
actual surge line. This operating margin is required because if the
80 compressor crosses over into unstable surging it cannot be returned
200.0 to stable operation by the closed-loop control system. A vibration
70
Pressure, psia

60 interlock system is often used to throw open the flow valve in order
to move the compressor out of its surge condition.
150.0

Anti-Surge
100.0
Control Line REFERENCES
T.L. Blevins et al., Advanced Control Unleashed, Research Triangle Park:
50.0 ISA—The Instrumentation, Systems, and Automation Society, 2003.
T.F. Edgar et al., Chapter 8, Process Control. Perry’s Chemical Engineering
Handbook, 6th ed., McGraw-Hill, New York, 1999.
0.0
W.L. Luyben, Practical Distillation Control, Van Nostrand Reinhold, New
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 York, 1992.
Flow Rate, kpph
G.K. McMillan, Process/Industrial Instruments and Controls Handbook,
5th ed., McGraw-Hill, New York, 1999.
G.K. McMillan, Tuning and Control Loop Performance, 3rd ed., Research
Figure 3.30. Compressor surge curve.
Triangle Park: ISA—The Instrument Society of America, 1994.
F.G. Shinskey, Feedback Controllers for the Process Industries Systems,
McGraw-Hill, New York, 1994.
these peaks may be connected to determine a surge line, which F.G. Shinskey, Process Control Systems, 3rd ed., McGraw-Hill, New York,
defines a region of unstable operation at lower flows. 1988.
Flow control by recycle of process gas is used to maintain the C.A. Smith and A.B. Corripio, Principles and Practice of Automatic Process
minimum flow requirements of the anti-surge control system. Control Systems, John Wiley & Sons, New York, 1985.