Professional Documents
Culture Documents
(Armando) Tuning of Industrial Control Systems
(Armando) Tuning of Industrial Control Systems
Third Edition
By
Armando B. Corripio, Ph.D., P.E.
Chemical Engineering
Louisiana State University
and
Michael Newell
Automation Designer
Polaris Engineering
Notice
The information presented in this publication is for the general education of the reader. Because
neither the author nor the publisher has any control over the use of the information by the reader,
both the author and the publisher disclaim any and all liability of any kind arising out of such use.
The reader is expected to exercise sound professional judgment in using any of the information
presented in a particular application.
Additionally, neither the author nor the publisher has investigated or considered the effect of
any patents on the ability of the reader to use any of the information in a particular application.
The reader is responsible for reviewing any possible patents that may affect any particular use of
the information presented.
Any references to commercial products in the work are cited as examples only. Neither the
author nor the publisher endorses any referenced commercial product. Any trademarks or
tradenames referenced belong to the respective owner of the mark or name. Neither the author
nor the publisher makes any representation regarding the availability of any referenced
commercial product at any time. The manufacturers instructions on the use of any commercial
product must be followed at all times, even if in conflict with the information in this publication.
ISBN: 978-0-87664-034-0
No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form
or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior
written permission of the publisher.
ISA
67 Alexander Drive
P.O. Box 12277
Research Triangle Park, NC 27709
This third edition of Tuning of Industrial Control Systems has been significantly
simplified from the second edition with the goal of having the discussion
more in line with modern control systems and with language that is less aca-
demic and more in tune with the vocabulary of the technicians who do the
actual tuning of control systems in industry. For example, we have eliminated
any references to first- and second-order models since these terms are highly
mathematical and may discourage some from appreciating the usefulness of
the models. We have also eliminated the distinction between series and paral-
lel PID controllers since most modern installations use the series version and
there is not much difference between the tuning of the two versions.
Chapter 10 is new and deals with the auto-tuning feature that has become
standard on current process control systems. We have successfully used the
auto-tuning feature in our tuning work on oil refineries as a reference to guide
our selection of the final tuning parameters for the controllers.
ix
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
v
vi Tuning of Industrial Control Systems, Third Edition
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
1
Introduction
This first chapter presents a general discussion of the goal of tuning, a descrip-
tion of feedback controlthe most common strategyand a brief introduc-
tion to other common control strategies.
Learning Objectives When you have completed this chapter, you should be
able to
1
2 Tuning of Industrial Control Systems, Third Edition
The following heuristics (rules of thumb) may prove helpful to those just
starting in the tuning of processes:
The controller cannot move the process variable faster than the process
can respond, so the controller speed must be matched to the speed of
response of the process. Some processes respond in a matter of min-
utes, while others may take close to an hour or longer to respond. Not
many processes respond in a matter of seconds.
One more item to keep in mind is that there is no such thing as fine-tuning a
controller, particularly a feedback controller. In most cases the tuning parame-
ters need only be adjusted to one, or at most, two significant digits. There are
two reasons for this. One is that feedback controllers are not that sensitive to
variations in the third digit of their tuning parameters. The other is that the
characteristics of most processesthat is, speed of response and sensitivity to
changes in controller outputvary with operating conditions, sometimes
slightly and other times not so slightly. This means that the controller tuning
Introduction 3
Understanding this simplifies the task of tuning because it reduces the num-
ber of values of the tuning parameters to be tried. For example, it is a lot easier
to decide between gain values of 1.0 or 1.5 than to try to find out whether the
gain should be 1.276. In practice, all three of these values will work about the
same.
Armed with these heuristics and basic concepts, we are now ready to look at
the feedback control strategy.
Feedback control is the basic strategy for the control of industrial processes. It
consists of measuring the process variable to be controlled (the controlled
variable), comparing the measurement with its desired value or set point, and
taking action based on the difference between them to reduce or eliminate the
differencethat is, to bring the controlled variable to its desired value. The
action taken results in the adjustment of a process flow, such as the steam flow
to a heater, which has a direct effect on the controlled variable, such as the
outlet process temperature. The three instrumentation components required
for feedback control are:
A fourth element of the loop is the process itself, through which the manipu-
lated flow affects the controlled variable. The controlled variable is also
known as the process variable (PV), its desired value is the set point (SP), and the
signal from the controller to the final control element is the controller output
(OP).
4 Tuning of Industrial Control Systems, Third Edition
It is important to realize that a feedback controller does not use a model of the
process to compute its output. It takes action by trial and error. Tuning the
controller is the procedure of adjusting the controller parameters to ensure
that the controller output converges quickly to its correct value.
SP
Steam
OP
TC
Fs
PV
F TT
Ti
Process T
fluid
Steam
trap
Condensate
Introduction 5
In Figure 1-1 the sensor transmitter is shown as a circle with the letters TT in it
and the feedback controller is a circle with the letters TC in it. This follows the
standard ISA instrumentation notation1 in which the first letter denotes the
variable being measured or controlled, in this case T for temperature, and
the second letter is T for the transmitter and C for the controller. The con-
trol valve is represented by the symbol shown on the steam line to the heater.
Its purpose is to adjust the flow of steam (Fs) in response the controller output
signal (OP).
The transmitter and the control valve are located in the field while the control-
ler is located in a central control room. Today, the signals between the trans-
mitter and the controller and between the controller and the control valve are
typically digital signals transmitted through a fieldbus or by wireless transmis-
sion. The control function is carried out by a computer or distributed control
system (DCS) that receives the transmitter signal and transmits the controller
output to the control valve. The control valve is usually pneumatically oper-
ated, requiring that the controller output be converted to an air pressure sig-
nal. This is done by a current-to-pressure (I/P) transducer.
Figure 1-2 is a block diagram of the feedback control loop for the process
heater. It graphically shows the loop around which signals travel: a change in
outlet temperature T causes a proportional change in the signal PV to the con-
troller; the summer (circle), a part of the controller, calculates the error E or
6 Tuning of Industrial Control Systems, Third Edition
deviation of the process variable from the set point SP and acts on this error
by changing the signal OP to the control valve; the control valve position
changes, causing a change in steam flow Fs to the heater; this in turn causes a
change in the outlet temperature T which then starts a new cycle of changes
around the loop.
Figure 1-2. Block Diagram of the Temperature Control Loop of the Process
Heater
F Ti
- +
SP + E + OP+ Fs + T
Controller Control Valve Heater
-
PV Sensor
+
Transmitter
The signs in Figure 1-2 represent the action of the various input signals on the
output signal; that is, a positive sign means that an increase in input causes an
increase in outputdirect actionwhile a negative sign means that an increase
in input causes a decrease in outputor reverse action. For example, the nega-
tive sign by the process flow into the heater means that an increase in flow
results in a decrease in outlet temperature. Notice that by following the sig-
nals around the loop, there is a net reverse action in the loop. This property is
known as negative feedback and is a required characteristic of a feedback loop
for the loop to be stable. In this example it means that an increase in heater
outlet temperature results in a decrease in controller output, which in turn
closes the control valve and reduces the steam flow. This results in a decrease
in outlet temperature, as desired.
To ensure this self-regulating effect the controller must act in the correct direc-
tion when the process variable changes. In this example the controller action is
reverse, that is, an increase in process variable results in a decrease in control-
Introduction 7
ler output. Other processes may require direct action, for example when a tank
level controller adjusts the flow out of the tank. In this case, an increase in liq-
uid level in the tank requires that the exit control valve open to increase the
flow out of the tank and decrease the level. Consequently, the action (direct or
reverse) of the feedback controller is its most important characteristic.
Although feedback control is by far the most common automatic control strat-
egy, there are other strategies that have been known to enhance control per-
formance in terms of improving loop stability, preventing initial deviation of
the process variable, and allowing tighter control. This section will briefly
introduce these strategies; their details and tuning procedures will be pre-
sented in later chapters.
the process when each controller output affects the process variables
controlled by the other controllers.
1-5. Summary
This first chapter has presented the goals of the tuning procedure and has
introduced the feedback control strategy. A brief description of other common
control strategies has also been presented.
References
Review Questions
1-2. Which two process characteristics must be considered when tuning the
controller?
The basic concept of feedback control was introduced in the preceding chap-
ter. This chapter presents details of the feedback controller and one of the
methods proposed to tune it: the ultimate gain and period method.
11
12 Tuning of Industrial Control Systems, Third Edition
The previous chapter showed that the purpose of the feedback controller is to
compute its output signal based on the difference between the controlled pro-
cess variable and its desired value or set point. This section presents the three
basic modes used by the controller to compute its output signal.
The three basic modes of feedback control are proportional (P), integral (I)also
called resetand derivative (D)also called rate. The controller can function
in a single mode or in a combination of either two modes or of all three,
although today most controllers function in either two or three modes. Either
way the device is known as a PID controller, based on the assumption that it
can function in all three modes. Each of these modes introduces an adjustable
or tuning parameter into the operation of the feedback controller.
Proportional Mode
The proportional mode does not eliminate the sustained deviation (off-
set) between the process variable PV and the set point SP.
Figure 2-1. Response of the Proportional Mode with the Loop Open
PV
SP
OP
Kc = 1.0
Kc = 2.0
time
Figure 2-2. Response of the Integral Mode to a Step Change in PV with the Loop
Open
PV
SP
Kc = 1.0
OP
TI = 6
TI = 12
time
The step in output shown in the figure is the instantaneous response of the
proportional mode. It takes the integral mode a period of time equal to TI to
duplicate the instantaneous response of the proportional mode.
The integral mode thus forces the process variable to the set point at the
expense of slower action than the proportional mode. This slow action intro-
duces some instability into the response of the loop.
The derivative mode action is zero when the process variable remains
constant.
Figure 2-3. Response of the Derivative Mode to a Ramp in the PV with the Loop
Open
PV
SP
OP
TD = 6.0
TD = 3.0
0
time
16 Tuning of Industrial Control Systems, Third Edition
To better illustrate the anticipation action of the derivative mode, the response
to a ramp in the process variable is shown in Figure 2-4 for a direct-acting con-
troller having both proportional mode (with a gain of 1.0) and derivative
mode. The initial step in the output is caused by the derivative mode and the
continuous change is caused by the proportional mode. As a result, the output
leads the process variable by a period of time equal to the derivative time.
Notice that this does not mean the controller can predict the future, since the
output cannot change until the process variable starts changing.
Kc = 1.0
OP
10
TD = 10 PV
SP
time
Although the derivative mode increases the stability of the control loop, it has
two undesirable characteristics. One is that if the transmitter signal (PV) is
noisy, the derivative can amplify noise. To limit this amplification as the fre-
quency of the noise increases, practical controllers have a built-in filter on the
derivative mode that limits the amplification factor. The other undesirable
characteristic is that the derivative mode can cause sudden changes in control-
ler output with sudden changes in the process variable. This is usually not a
problem because very seldom will the process variable change suddenly in
practice. To prevent sudden changes in set point from causing sudden
changes in output, all practical controllers have the derivative mode work
only on the process variable, not on the deviation from the set point.
The Feedback Controller 17
The three adjustable tuning parameters of the PID controller are the propor-
tional gain Kc, the integral time TI, and the derivative time TD. The time
parameters are specified in minutes for most controllers, although some
brands may require them in seconds. Although modern control systems dis-
play the process variable in engineering units (F, lb/hr, barrels/day, psi,
etc.), the proportional gain is dimensionless, because it is defined as the
change in percent controller output per percent change in the process vari-
ables transmitter output (i.e., the fraction, in percent, that the process variable
value is of the calibrated range of the transmitter).
Figure 2-5 illustrates this concept for a temperature controller. The left scale
shows the process variable PV both in engineering units, F, and percent of
transmitter output. The transmitter is calibrated to measure the temperature
in the range of 50F (0% of the range) to 250F (100% of the range). The set
point SP is assumed to be in the middle of the range, 150F or 50%.
Figure 2-5. Process Variable in Engineering Units (F) and Percent of Range.
Illustration of Controller Proportional Band
T, F PV, % OP, %
250 100 100
80 80
60 60
20 20
50 0 0
18 Tuning of Industrial Control Systems, Third Edition
Figure 2-5 also illustrates the concept of the controller proportional band (PB)
defined as the fraction of the transmitter output range that causes a 100%
change in the controller output OP. For the assumed gain of 5.0 the propor-
tional band is 20%. In some older controllers the gain was specified as the pro-
portional band, but that is no longer the practice.
The feedback controllers are displayed for the operators in the control console
and provide the following features:
Auto/Manual switch
With these the operator can observe the current value of all the variables asso-
ciated with the control loop, adjust the set point, and if necessary switch the
controller to Manual and adjust the controller output. In cascade control sys-
tems the operator can switch the slave controller to local set point and
adjust its set point. The controllers are programmed so that the switching
from Manual to Auto and from local to remote set point is bump-less; that is,
The Feedback Controller 19
the controller output does not change, and, optionally, the set point is set to
the current value of the process variable when the switch is performed.
Having introduced the feedback controller in this section, the next section
presents the concept of loop stability, that is, the effect of the controller on the
process response.
One of the characteristics of feedback control loops is that they may become
unstable. The loop is said to be unstable when a small change in a disturbance
variable or the set point causes the system to deviate widely from its normal
operating point. The two possible causes of instability are that the controller
has the incorrect action (direct when it should be reverse or vice versa) or that
it is tuned too tightlythat is, the gain is too high, the integral time is too
short, the derivative time is too long, or a combination of these. Another possi-
ble cause is that the process is inherently unstable, but this is rare.
When the controller has the incorrect action, instability is manifested by the
controller output running away to either its upper or its lower limit. For
example, if the temperature controller on the process heater of Figure 1-1 were
set so that an increasing temperature increases its output, a small increase in
temperature would result in an opening of the steam valve, which in turn
would increase the temperature some more and the cycle would continue
until the controller output was at its maximum with the steam valve fully
open. On the other hand, a small decrease in temperature would result in a
closing of the steam valve, which would further reduce the temperature, and
the cycle would continue until the controller output was at its minimum point
with the steam valve fully closed. Thus, the stability of the temperature con-
trol loop of Figure 1-1 requires that the controller decrease its output when the
process variable increases. As we have seen, this is known as reverse action.
20 Tuning of Industrial Control Systems, Third Edition
As pointed out earlier, the oscillatory type of instability is caused by the con-
troller having too high a gain, too short an integral time, or too long a deriva-
tive time, or a combination of these. This leads into the simplest method for
characterizing the process for the purpose of tuning the controller, that of
determining the ultimate gain and period of oscillation of the loop.
The Feedback Controller 21
The first controller tuning method will now be introduced, one that depends
on measuring the characteristics of the control loop by determining the limit
of stability of the closed loop with a proportional controller.
The earliest published method of characterizing the process for controller tun-
ing was proposed by J. G. Ziegler and N. B. Nichols.1 The method consists of
determining the ultimate gain and period of oscillation of the loop. The ulti-
mate gain is the gain of a proportional controller at which the loop oscillates
with constant amplitude, and the ultimate period is the period of the oscilla-
tions. The ultimate gain is thus a measure of the controllability of the loop;
that is, the higher the ultimate gain, the easier it is to control the loop. The ulti-
mate period is in turn a measure of the speed of response of the loop, that is,
the longer the period, the slower the loop.
It follows from the definition of the ultimate gain that it is the gain at which
the loop is at the threshold of instability. At gains higher than the ultimate
gain, the loop signals oscillate with increasing amplitude, as in Figure 2-6. Fig-
ure 2-7 shows the response of a loop to a disturbance (for example, an increase
in process flow in the heater of Figure 1-1) with a proportional controller at
increasing values of the controller gain. As the figure shows, as long as the
gain is lower than the ultimate gain, the amplitude of the oscillations
decreases with time. When determining the ultimate gain it is very important
to approach it in small gain increments to ensure that it is not exceeded by
much, lest the system become violently unstable.
The procedure for determining the ultimate gain and period is carried out
with the controller in Automatic and with the integral and derivative modes
removed. It is as follows:
1. Remove the integral mode by setting the integral time to its highest value.
Alternatively, if the controller model or program allows for switching off
the integral mode, switch it off.
2. Switch off the derivative mode or set the derivative time to its lowest
value, usually zero.
22 Tuning of Industrial Control Systems, Third Edition
Kc = 2
Kc = 0.5
Kc = 2
Kc = 0.5
3. Carefully increase the proportional gain in steps. After each increase, dis-
turb the loop by introducing a small step change in set point and observe
the response of the controlled and manipulated variables, preferably on a
trend recorder. The variables should start oscillating as the gain is
increased, as in Figure 2-7.
5. Measure the period of the oscillations from the trend recordings, as in Fig-
ure 2-8. For better accuracy, time several oscillations and calculate the
average period. In Figure 2-8, for example, the time required by five oscil-
lations is measured and then divided by 5.
6. Stop the oscillations by reducing the gain to about half of the ultimate
gain.
The Feedback Controller 23
Kcu = 3.42
5Tu
The procedure just outlined is simple and requires a minimum upset to the
process, just enough to be able to observe the oscillations. Nevertheless, the
prospect of taking a process control loop to the verge of instability is not an
attractive one from a process operation standpoint. However, it is not abso-
lutely necessary in practice to obtain sustained oscillations (see the section on
Practical Ultimate Gain Tuning Tips). It is also important to realize that some
simple loops cannot be made to oscillate with constant amplitude with just a
proportional controller. Fortunately, these are usually the simplest loops to
control and tune.
Along with the method just outlined for determining the ultimate gain and
period of a feedback control loop, Ziegler and Nichols proposed a set of for-
mulas to tune the feedback controller for a specific response, the quarter-
decay-ratio response, or QDR for short. The QDR response is illustrated in
Figure 2-9 for a step change in set point and for a step change in disturbance.
24 Tuning of Industrial Control Systems, Third Edition
A/4
SP
SP
A/4
Disturbance
A
The Feedback Controller 25
It is intuitively obvious that for the proportional (P) controller the gain for a
QDR response should be one-half of the ultimate gain, as Table 2-1 shows. At
the ultimate gain, the maximum error in each direction causes an identical
maximum error in the opposite direction; at one-half the ultimate gain, the
maximum error in each direction is exactly one-half the preceding maximum
error in the opposite direction and one-fourth the previous maximum error in
the same direction. This is the quarter-decay-ratio response.
In Table 2-1 notice that the addition of integral mode results in a reduction of
10% in the QDR gain between the P and the PI controller tuning formulas.
This is due to the additional lag introduced by the integral mode. On the other
hand, the addition of the derivative mode allows increasing the controller
gain by 20% over the proportional controller. Therein lies the justification for
the derivative mode, that is, the increase in the controllability of the loop.
Finally, the derivative and integral times in the PID formulas are in the ratio of
1:4. This is a useful relationship to keep in mind when tuning PID controllers
by trial-and-error (i.e., in those cases when the ultimate gain and period can-
not be determined).
Determine the ultimate gain and period for the temperature control loop of
Figure 1-1, and the quarter-decay tuning parameters for a P, a PI, and a PID
controller.
For the temperature control loop, Figure 2-8 shows responses of the process
variable PV and the controller output OP with a proportional controller and
a gain of 3.42, which results in sustained oscillations. The ultimate gain is
then 3.42. A small change in the flow to the heater is used to start the oscilla-
tions. In the figure, the period of the oscillations is the ultimate period.
37.5 3.5
Ultimate period: ------------------------ = 6.8 min
5
26 Tuning of Industrial Control Systems, Third Edition
Using the formulas of Table 2-1, the QDR tuning parameters are:
Figure 2-10 shows the response of the controller output and the outlet process
temperature to an increase in process flow for the proportional controller with
a QDR gain of 1.7 and with a lower gain of 1.0. The figure shows that the
lower gain results in a larger initial deviation of the PV and a larger offset, but
the oscillations are smaller and the required variation in controller output is
less.
Kc = 1.7
Kc = 1.0
Kc = 1.7
Kc = 1.0
The Feedback Controller 27
Figures 2-11 and 2-12 show the responses of the PI and PID controllers,
respectively. In each case, the smaller proportional gain results in less oscilla-
tory behavior and less initial movement of the controller output, at the
expense of a larger initial deviation of the PV and a slower return to the set
point. This shows that the tuning parameters, particularly the gain, can be
varied from the values given by the tuning formulas to obtain the desired
response.
Kc = 1.5
Kc = 1.0
TI = 5.7 min
Kc = 1.5
Kc = 1.0
Notice the offset in Figure 2 10, and the significant improvement that the
derivative mode produces in the responses of Figure 2-12 over those of
Figure 2-11.
Kc = 2.0
Kc = 2.0
Kc = 1.0
period gives good enough values of the integral and derivative times. The
proportional gain can then be adjusted to obtain an acceptable response.
For example, notice in Figure 2 7 that, for the case of a gain of 2, the period
of oscillation is 8.0 minutes, which is less than 20% away from the actual
ultimate period (6.8 min).
The QDR tuning formulas allow the tuning of controllers for a specific
response when the ultimate gain and period of the loop can be determined.
The Feedback Controller 29
The chapters that follow present alternative methods for characterizing the
dynamic response of the loop and for tuning feedback controllers. The follow-
ing section brings up the need for such alternative methods.
Although the ultimate gain tuning method is simple and fast, other methods
of characterizing the dynamic response of feedback control loops have been
developed over the years. The need for these alternative methods is based on
the fact that it is not always possible to determine the ultimate gain and period
of a loop. As pointed out earlier, some simple loops will not exhibit constant
amplitude oscillations with a proportional controller.
The ultimate gain and period, although sufficient to tune most loops, do not
give insight into which process or control system characteristics could be
modified to improve feedback controller performance. A more fundamental
method of characterizing process dynamics is needed to guide such
modifications.
There is also the need to develop tuning formulas for responses other than the
quarter-decay-ratio response. This is because the set of PI and PID tuning
parameters that produce a quarter-decay response are not unique. It is easy to
see that for each setting of the integral and derivative time, there will usually
be a setting of the controller gain that will produce a quarter-decay response.
This makes for an infinite number of combinations of the tuning parameters
that satisfy the quarter-decay-ratio specification.
2-5. Summary
References
Review Questions
2-1. A controller has a gain of 3. For each of the following cases determine by
how much the proportional mode causes the output of the controller to
change and in which direction increase or decrease:
2-4. A controller is switched to Automatic and its output starts rising imme-
diately and does not stop until it reaches its upper limit. What do you
think is the cause?
2-6. Why do you think that the tuning formulas of Table 2-1 relate the inte-
gral and derivative times to the ultimate period of oscillation of the
loop?
2-7. After tuning a controller using the formulas of Table 2-1 you find the
variation in the controller output when a disturbance upsets the system
is higher than you would like it to be. How would you adjust the tuning
to obtain a more reasonable behavior?
3
Open-loop
Characterization of
Process Dynamics
33
34 Tuning of Industrial Control Systems, Third Edition
The preceding chapter showed how to determine the ultimate gain and period
of a feedback control loop by performing a test with the controller on auto-
matic output (i.e., with the loop closed). By contrast, this chapter shows how
to determine the process parametersgain, time constant, and dead timeby
performing a test with the controller on manual output, that is, an open-
loop test. Such a test presents a more fundamental model of the process than
the ultimate gain and period.
SP
Steam
OP
TC
Fs
PV
F TT
Ti
Process T
fluid
Steam
trap
Condensate
The simplest type of open-loop test is a step testthat is, a sudden and sus-
tained change in the controller output OP. Figure 3-2 shows a typical step test
and the response of the process variable. The S-shaped response is typical of
most processes that are self-regulating (i.e., they reach a steady value after the
response time is over). What causes this type of response is the fact that the
outputs of the different components of the control loop lag their inputs in time
(i.e., their outputs do not immediately respond to their inputs). So there are
lags in the control valve and the sensor/transmitter, as well as one or more
lags in the process. Because of this the process variable does not start chang-
ing right after the step change is applied. As Figure 3-2 shows, the rate of
change starts at zero and then increases to a maximum rate that is followed by
a decreasing rate as the variable approaches its final steady value.
36 Tuning of Industrial Control Systems, Third Edition
Another reason that generally the process variable does not start changing
immediately is the presence of transportation lag in the loop. This is the lag
caused by the time it takes for the process fluid to move through the process.
However, for most loops real transportation lag, usually of the order of sec-
onds, is negligible relative to the lags that are commonly of the order of min-
utes. (See Section 3-5 for further discussion of transportation lag.)
This section shows how to extract the process characteristic parameters from
the results of a step test using the step test of Figure 3-2 as an example. The
three parameters of interest to the tuning of the feedback controller are:
The process time constant, a measure of how long it takes for the process
to respond.
The process dead time, a measure of how long it takes for the process
variable to start changing after the step change in controller output is
applied.
These three parameters are discussed in detail below. Together they constitute
a simple-lag-plus-dead-time (SLPDT) model of the loop response that will be
used to tune the feedback controller.
These parameters are estimated from the step response of the loop and are
then used to tune the feedback controller.
Gain
The gain determined from the step response is the product of the gains of the
control valve (or other final control element), the process, and the sensor/
transmitter. It must be expressed as percent change in transmitter output per
percent change in controller output, not in engineering units. Figure 3-3
shows the determination of the gain from the step response of Figure 3-2. As
in most control computer-generated plots, the PV is displayed in engineering
units, F, and the final steady change in the PV is (209.5 190) = 19.5F. To
convert to percent of the transmitter output, we need the range of the trans-
mitter. Let us say it is 50 to 250F; the change in the PV is then:
19.5F
-------------------------------- 100% = 9.75%
( 250 50 )F
Since this change is caused by a 5% change in controller output, the gain is:
9.75%
--------------- = 1.95
5%
38 Tuning of Industrial Control Systems, Third Edition
t2 = 15 5 = 10 min
190 + 0.632(19.5) =
202.3F
209.5 190 = 19.5F
190 + 0.283(19.5) = 195.5F
t1 = 10 5 = 5 min
The dynamic response is characterized by two parameters: the dead time and
the time constant. Although Ziegler and Nichols2 proposed two different
parameters that must be determined by graphical construction, Cecil Smith1
proposed to fit a model consisting of dead time and a single lag to the
response. Smiths method requires only reading two points from the response
and avoids the cumbersome graphical construction. The procedure consists of
reading the times in the response at which the PV reaches 28.3% and 63.2% of
its total change. These times correspond, respectively, to one-third of the time
constant and one time constant in the response of a single lag, so the differ-
ence between them is two-thirds of the time constant. They are picked in the
area of high rate of change of the response, which results in more accurate
determination of the two time values. The dynamic parameters are then:
where t1 and t2 are, respectively, the times at which the response reaches
28.3% and 63.2% of its total change.
Note that we subtracted 5 minutes from the times read in the response to
determine the times from the application of the step change in controller
output.
Figure 3-4 shows the step response of a hot oil temperature controller in a
refinery. The coil outlet temperature is controlled by manipulating the flow
of hot oil through the coil, which is in the convection section of a furnace. A
step increase of 3% in the controller output at time zero results in the
response shown in the figure. The response data are read off the trend plot
and entered into a spreadsheet for plotting and analysis. The range of the
temperature transmitter is 0 to 1200F. The change in the process variable is:
( 635.8 638.2 )F
-------------------------------------------- 100% = 0.8%
( 1200 0 )F
Figure 3-4. Determination of Loop Parameters from Step Test of a Hot Oil
Temperature Controller
638.5
638.2F
638.0
637.0
636.0
t2 = 87 sec
635.8F
635.5
0 50 100 150 200 250 300 350 400
time, sec
Measurement Calculated
From the figure, t1 = 54 seconds and t2 = 87 seconds. The data were manually
obtained from the trend recorder on the control system in a refinery and
entered into a spreadsheet. Interpolation was then used on the spreadsheet to
obtain the model parameters. The time constant and dead time are then:
The dashed line in Figure 3-4 is a plot of the response of a process with a gain
of 0.267, a transportation lag of 37 sec, and a time constant of 50 sec.
The negative sign of the gain means that an increase in controller output
results in a decrease in temperature because it causes an increase in the pro-
cess flow through the coil. The controller must then have direct action.
Open-loop Characterization of Process Dynamics 41
This example illustrates how modern computer control systems allow the pre-
cise determination of the step response parameters using a small step change
in controller output that results in a very small change in the process variable
(less than 1%).
There are two types of processes that are not self-regulating: imbalanced or
integrating processes, and open-loop unstable processes. A typical example of
an imbalanced process is the liquid level in a tank, and an example of an
unstable process is an exothermic chemical reactor. It is obviously impractical
to perform step tests on processes that are not self-regulating. Fortunately,
most processes are self-regulating.
Although the process time constant and dead time can be estimated from an
open-loop step test (as described in the previous section), it is important to
examine the physical significance of these two dynamic parameters of the pro-
cess. Doing this will allow estimation of the process time constant and dead
time from physical process characteristics (e.g., volumes, flow rates, valve
sizes) when it is not convenient to perform the step test. This section concerns
the time constant, and the next section explores the dead time.
Figure 3-5. Physical Systems with Simple Lag Dynamics: (a) Electrical Circuit; (b)
Liquid Storage Tank; (c) Gas Surge Tank; (d) Blending Tank
R
Inlet flow
ein C eout
Level
Outlet flow Kv
(b)
(a)
Outlet flow F1 C1 F C
Inlet flow
P V T F2 C2 V
Kv
(c) (d)
The time constant of a lag is defined as the ratio of its capacitance to its con-
ductance or the product of the capacitance times the resistance (the resistance
is the reciprocal of the conductance):
Capacitance
= ---------------------------------- = Capacitance Resistance
Conductance
The conductance is the ratio of the flow to the potential that drives it:
Flow of quantity conserved
Conductance = -------------------------------------------------------------------------
Potential
To better understand the physical meanings to the terms just presented, con-
sider each of the physical systems of Figure 3-5.
Electrical Circuit (Figure 3-5a). For this system the quantity conserved is elec-
tric charge, the potential is electric voltage, and the flow is electric current. The
capacitance is provided by the ability of the capacitor to store electric charge
and the conductance is the reciprocal of the resistance of the electrical resistor.
The time constant is then given by:
= RC
where:
Liquid Storage Tank (Figure 3-5b). In this common process system, the quan-
tity conserved is the volume of liquid (assuming constant density), the capaci-
tance is provided by the ability of the tank to store liquid, and the potential for
flow through the valve is provided by the level of liquid in the tank. The
capacitance is volume of liquid per unit level (i.e., the cross-sectional area of
the tank), and the conductance is the change in flow through the valve per
unit change in level. The time constant can then be estimated by:
= A/Kv (3-1)
where:
The conductance of the valve depends on the valve size, and is usually known
in terms of flow per unit pressure drop. Note that the change in pressure drop
across the valve per unit change in level can be calculated by multiplying the
density of the liquid times the local acceleration of gravity.
Gas Surge Tank (Figure 3-5c). This system is analogous to the liquid storage
tank. The quantity conserved is the mass of gas, the potential that drives the
flow through the valve is the pressure in the tank, and the capacitance is pro-
vided by the ability of the tank to store gas as it is compressed. The capaci-
tance can be calculated by the formula MV/zRT lb/psi, where V is the volume
of the tank (ft3), R is the ideal gas constant (10.73 psi ft3/lbmole R), z is the
compressibility factor of the gas, M is its molecular weight (lb/lbmole), and T
is its absolute temperature (R). The conductance of the valve is expressed in
change of mass flow per unit change in pressure drop across the valve. The
time constant of the tank can be estimated by the formula:
= (MV/zRT)/Kv (3-2)
where:
= V/F (3-3)
where:
If there is a chemical reaction, the time constant for the concentration of reac-
tants is decreased because the conductance is increased to the sum (F + kV)
where k is the reaction coefficient, defined here as the change in reaction rate
divided by the change in the reactant concentration. The conductances are
added because the processes of reaction and convection occur in parallel.
For the preceding examples of simple lags the time constant may be estimated
from process parameters and thus a dynamic test on the process is not
needed. For more complex processes such as distillation columns and heat
exchangers, the time constant cannot be estimated because it is made up of
many resistance-capacitance combinations in series and in parallel. For these
systems the only recourse is to perform a dynamic test such as the one pre-
sented earlier in this chapter.
The surge tank of Figure 3-5c is for an air compressor. It runs at a tempera-
ture of 150F, and has a volume of 10 ft3. The valve can pass a flow of
100 lb/hr at a pressure drop of 5 psi when the pressure in the tank is 30 psig.
Estimate the time constant of the response of the pressure in the tank to
variations in inlet pressure.
The capacitance of the tank is its ability to store air as the density of air
changes with pressure, which is the potential for flow. Assuming that air at
30 psig behaves as an ideal gas (z = 1), and using the fact that its molecular
weight M is 29, the capacitance is:
The conductance of the valve can be estimated from the formulas given by
valve manufacturers to size the valves. Since the pressure drop through the
valve is small compared to the pressure in the tank, the flow is subcritical
and the conductance is given by the following formula:
Kv = W (1 + Pv/P)/(2Pv)
= (100/60)[1 + 5/(30+14.7)]/[(2)(5)] = 0.1853 (lb/min)/psi
The conductance calculated for the valve is the change in gas flow per unit
change in tank pressure, P. It takes into account the variation in gas density
with pressure, and the variation in flow with the square root of the product
of density times the pressure drop across the valve, Pv. For critical flow,
when the pressure drop across the valve is more than one-half the upstream
absolute pressure, the conductance can be calculated by the formula
Kv = W/P.
Pure dead time, also called transportation lag or time delay, occurs when the
process variable is transported from one point to another, hence the term
transportation lag. At any point in time, the variable downstream is what
the variable upstream was one dead time earlier, hence the term time delay.
This is all illustrated in Figure 3-6. When the upstream variable C1 first starts
changing at the upstream point, it takes one dead time before the downstream
variable C2 starts changing, hence the term dead time. The dead time can be
estimated from the following formula:
Distance L
t 0 = ---------------------- = --- (3-4)
Velocity v
Electric voltage and current travel at the velocity of light: 300,000 km/s
or 984,000,000 ft/s.
Open-loop Characterization of Process Dynamics 47
Figure 3-6. Physical Occurrence of Dead Time (Transportation Lag or Time Delay)
and Response
C1 C2
v
C1
Dead time
L/v
C2
Time
Pressure and flow travel at the velocity of sound in the fluid; for exam-
ple, 340 m/s or 1,100 ft/s for air at ambient temperature.
Solid properties vary at the velocity of the solid (e.g., paper in a paper
machine or coal in a conveyor).
These numbers show that for the reasonable distances, which are typical of
process control systems, pure dead time is only significant for temperature,
composition, and other fluid and solid properties. The velocity of a fluid in a
pipe can be calculated by the following formula:
v = F/Ap (3-5)
48 Tuning of Industrial Control Systems, Third Edition
where
Given that (as shall be seen shortly) the dead time makes a feedback loop less
controllable, most process control loops are designed to reduce the dead time
as much as possible. Dead time can be reduced by installing the sensor as
close to the equipment as possible or in the equipment itself.
Although pure dead time is usually not significant for most processes, the
process dead time estimated from the response to the step test arises from
phenomena which are not necessarily transportation lag, but consist of the
presence of two or more lags in series (e.g., the trays in a distillation column).
When these processes are modeled with a simple lag, the dead time is needed
to represent the delay caused by the multiple lags in series. Figure 3-7 shows
the response of composition in a blending train when it consists of one, three,
and five tanks in series, assuming that the total blending volume is the same.
For example, each of the five tanks has one-fifth the volume of the single tank.
As the figure shows, the higher the number of tanks in series, the longer it
takes for the process to start changing and the shorter the total response time.
This behavior makes the model dead-time-to-time-constant ratio higher, mak-
ing the loop less controllable since it takes the feedback controller longer to
see the change in PV relative to the time it takes the PV to respond.
Most real processes fall somewhere between the two extremes of single per-
fectly mixed processes, and transportation (unmixed) processes. The simple-
lag-plus-dead-time (SLPDT) model used to model such processes is the sim-
plest model that can be used for characterizing them. It is the model com-
monly used to tune the controllers by practitioners in industry and by auto-
tune software.
Open-loop Characterization of Process Dynamics 49
F V PV
V PV
1
F V/3 V/3 V/3
/
PV
F V/5 V/5 V/5 V/5 V/5
3
Ft/V
The variation of the process gain, time constant, and dead time with process
operating conditions causes the controller performance to vary as process con-
ditions change. Because of this, a controller is usually tuned so that its perfor-
mance is best at the design operating point and is acceptable over the
expected range of operating conditions.
The formulas in the preceding section show that for concentration and tem-
perature (Equation 3-3), the time constant is inversely proportional to the flow
and thus to the throughput. For liquid level and gas pressure (Equations 3-1
and 3-2), the time constant varies with the valve conductance, Kv, which var-
ies with the valve characteristics and the pressure drop across the valve, and
the dead time is inversely proportional to the velocity (Equation 3-4), which is
in turn proportional to the flow (Equation 3-5). Control valve characteristics
such as equal percentage are usually selected to maintain the process gain
constant, which for liquid level and gas pressure is equivalent to keeping the
valve conductance constant (the valve gain is the reciprocal of the valve
conductance).
100%
90%
80%
70%
60%
Valve capacity
50%
40%
30%
20%
10%
0%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Valve position
the controller output, provided that the pressure drop across the valve
remains approximately constant.
Reset windup is more common in batch processes and during the startup and
shutdown of continuous processes, but the possibility of windup must always
be kept in mind when tuning controllers. Some apparent tuning problems are
really caused by unexpected reset windup. Chapter 4 looks at reset windup in
more detail.
52 Tuning of Industrial Control Systems, Third Edition
The following example illustrates the variation of the process gain in a process
heater. It takes advantage of the fact that for the heater, the gain can be calcu-
lated from a simple steady-state energy balance on the heater.
At design conditions, the process flow through the heater of Figure 3-1 is
F=16 kg/s, process fluid inlet temperature is Ti=50C, and it is desired to
heat the process fluid to T = 90C. The process fluid has a specific heat of
Cp=3.75 kJ/kgC, and the steam supplies Hv = 2250 kJ/kg upon condensing.
Heat losses to the surroundings can be neglected. The temperature trans-
mitter range is 50 to 150C, the control valve has linear characteristic with
constant pressure drop, and delivers 2.0 kg/s of steam when fully opened.
Calculate the gain of the heater in terms of the sensitivity of the outlet tem-
perature to changes in steam flow.
An energy balance on the heater, neglecting heat losses, yields the follow-
ing formula:
where Fs is the steam flow and the other terms have been defined in the
statement of the problem. The desired gain is the steady state change in out-
let temperature per unit change in steam flow:
Notice that the gain is inversely proportional to the process flow F. At the
design flow of 16 kg/s the process gain is:
2250 kJ/kg C
---------------------------------------------- = 37.5 --------------
16kg 3.75kJ kg s
------------- --------------------
s kg * C
Open-loop Characterization of Process Dynamics 53
For the linear valve with constant pressure drop, the gain of the valve is
equal to its capacity, (2 kg/s)/100% = 0.02 kg/s/%, and the gain of the trans-
mitter is (100%)/(150C 50C) = 1.0%/C. The dimensionless gain is then:
C 0.020kg s %
K = 37.5 -------------- ---------------------------1.0 ------- = 0.75
kg s % C
Now, if the process were to be run at one-half its full capacity, 8 kg/s, the
gain at this capacity would be:
2250kJ kg 0.020kg s %
K = ------------------------------------------- ---------------------------1.0 ------- = 1.50
8kg 3.75kJ % C
---------- --------------------
s kg * C
This doubling of the process gain could cause the loop to become unstable
if the controller was tuned at the design process flow.
This example shows the variation of the process gain, indicating that the
process heater is nonlinear. As mentioned earlier, this decrease in process
gain with an increase in flow is characteristic of many process control sys-
tems, hence the popularity of equal-percentage control valves, which
exactly compensate for this gain variation.
3-6. Summary
This chapter showed how to perform and analyze a process step test to deter-
mine the parameters of a simple-lag-plus-dead-time (SLPDT) model of the
process. These parameters are the gain, the time constant, and the dead time.
It also presented the physical significance of these parameters and showed
how to estimate them from process design parameters for some simple pro-
cess loops. The chapters to follow will use the estimated dynamic parameters
to design and tune feedback, feedforward, and multivariable controllers.
54 Tuning of Industrial Control Systems, Third Edition
References
2. Ziegler, J. G., and Nichols, N. B., Optimum Settings for Automatic Con-
trollers, Transactions ASME, V. 64, Nov. 1942, p. 759.
Review Questions
3-3. Figure 3-9 shows the response of the composition out of a reactor to a
step change in the controller output at time 1.0 min. The composition
controller has a range of 0 to 1.5 lb/gal. Estimate the parameters of a sim-
ple-lag-plus-dead-time model of the response.
3-4. A passive low-pass filter can be built with a resistor and capacitor. For
use in printed circuit boards, the maximum magnitudes of these compo-
nents are, respectively, 10 megohms (million ohms) and 100 microfarads
(millionth of farad). What would be the maximum time constant of a fil-
ter built with these components?
3-5. The liquid surge tank of Figure 3-5b has an area of 20 ft2 and the valve
has a conductance of 50 gpm/ft of level change (1 ft3 = 7.48 gallons). Esti-
mate the time constant of the response of the level to a step change in
inlet flow.
3-6. The blending tank of Figure 3-5d has a volume of 2000 gallons. Calculate
the time constant of the composition response for product flows of (a) 50
gpm, (b) 500 gpm, and (c) 5000 gpm.
Open-loop Characterization of Process Dynamics 55
3-7. The blending tank of Figure 3-5d mixes 100 gpm of concentrated solution
at 20 lb/gallon with 400 gpm of dilute solution at 2 lb/gallon. Calculate
the steady-state product concentration in lb/gallon. How much would
the outlet concentration change if the concentrated solution rate were to
change to 110 gpm, with all other conditions remaining the same? Calcu-
late the process gain.
3-8. Repeat the previous question assuming that the initial rates are 10 gpm
of concentrated solution and 40 gpm of dilute solution, and the concen-
trated solution is changed to 11 gpm to do the test. Also estimate the time
constant of the tank for both questions if the tank has a volume of
5,000 gal.
4
How to Tune Feedback
Controllers
This chapter presents formulas for tuning controllers based on the three
parameters obtained from the open-loop step test presented in the previous
chapter: gain, time constant, and dead time.
Learning Objectives When you have completed this chapter, you should be
able to:
57
58 Tuning of Industrial Control Systems, Third Edition
Besides the formulas for quarter-decay ratio (QDR) response tuning based on
the ultimate gain and period of the loop (see Table 2-1 in Chapter 2), Ziegler
and Nichols3 also developed formulas for tuning feedback controllers for
QDR response which are based on process gain K, time constant , and dead
time to. These formulas are given in Table 4-1.
The formulas of Table 4-1 are very similar to those of Table 2-1. Notice, for
example, that in both sets of formulas the proportional gain of the PI control-
ler is 10% lower and the PID gain is 20% higher than that of the P controller,
and that the derivative or rate time is one-fourth of the integral or reset time
for the PID controller. The ratio of the integral time of the PID controller to
that of the PI controller is 1.7, which is also the same as in Table 2-1; that is, the
derivative mode allows the integral mode to be 1.7 times faster.
The formulas of Table 4-1, however, provide important insight into how the
parameters of the process affect the tuning of the controller and thus the per-
formance of the loop, in particular:
The controller gain must be reduced when the ratio of the loop dead
time to its time constant increases. This means that the controllability of
the loop decreases when the ratio of the process dead time to its time
constant increases, and leads us to define the ratio of dead time to time
constant as the uncontrollability parameter of the loop:
t
P u = ---0-
where:
Note that it is the ratio of the dead time to the time constant that deter-
mines the degree of uncontrollability of the loop. In other words, a pro-
cess with a long dead time is not uncontrollable if its time constant is
much longer than its dead time.
These three conclusions can be helpful in guiding the tuning of feedback con-
trollers, even in cases when the tuning formulas cannot be used directly
because the process parameters cannot be accurately estimated. For example,
if the performance of a well-tuned controller should deteriorate during opera-
tion, look for a change in the process gain, its uncontrollability parameter, or
its speed of response. At other times the controller performance may be poor
because the integral time is much shorter than the process response time,
because in such a case the process cannot respond as fast as the controller
wants it to respond. The point here is that the speed of response of the process
must be considered when setting the integral time. This is what the tuning for-
mulas of Tables 2-1 and 4-1 do.
The conclusions just drawn from the tuning formulas, coupled with the meth-
ods for estimating time constants and dead times given in Sections 3-4 and 3-
5, can also guide the design of the process and its instrumentation. For exam-
60 Tuning of Industrial Control Systems, Third Edition
ple, loop controllability can be improved by reducing the dead time between
the manipulated variable and the sensor, or by increasing the process time
constant. It is also possible to quantitatively estimate the effect of process, con-
trol valve, and sensor nonlinearities on the variability of the loop gain and
thus determine the need for readjusting the controller gain when process con-
ditions change.
Although a number of sets of tuning formulas have been developed since the
Ziegler-Nichols3 formulas, all of them give controller parameters in the same
ballpark and in tuning, ballpark values are all that is needed. The Ziegler-
Nichols formulas of Table 4-1 are thus sufficient to get the tuning procedure
started. However, these formulas are empirical, and can be expected to be
valid for a common range of loop controllability, with say uncontrollability
values of 0.05 to 0.5. A set of tuning formulas based on more fundamental the-
ory was developed by Martin1, and independently by Rivera and Morari2.
These formulas, now commonly known as the Internal Model Control (IMC)
formulas, give results similar to the QDR formulas except for the integral
time, which is set equal to the loop time constant:
TI =
As expected, this results in more conservative (slower) tuning than with the
QDR formulas when the loop uncontrollability parameter is less than 0.5 and
faster responses for loops with higher dead-time-to-time-constant ratios.
Therefore, the following strategy is indicated:
Use the QDR formulas but limit the integral time to a maximum value equal
to the model time constant.
The IMC tuning method does not result in a specific value for the controller
gain since the formula for it contains an adjustable parameter, so it is just as
good to use the gain from the QDR formulas. In addition, the formula for the
derivative time for the IMC method is exactly the same as the one for QDR.
In Section 3-2, it was determined that the open-loop test parameters for the
heat exchanger of Figure 3-1 were:
Gain = 1.95
Time constant = 7.5 min
Dead time = 2.5 min.
The formulas of Table 4-1 produce the following QDR tuning parameters
for a PI controller:
The QDR parameters for a PID controller, also from Table 4-1, are:
Figure 4-1 compares the PI and PID controller responses of the temperature
transmitter output PV and of the controller output OP using these tuning
parameters for a step increase in process flow to the heater. For this loop with
an uncontrollability parameter of 2.5/7.5 = 0.33, the advantage of adding the
derivative mode is obvious: it produces a smaller initial deviation and main-
tains the temperature closer to the set point for the entire response, with fewer
oscillations. In addition, the initial change in controller output is not much
greater for the PID response than it is for the PI response.
The next two examples address the question of whether the QDR tuning
parameters will always perform this well, regardless of the degree of loop
controllability.
62 Tuning of Industrial Control Systems, Third Edition
PID
PI
PID
PI
An open-loop test on a loop results in a time constant of 11.6 min and a dead
time of 0.9 min. Since the uncontrollability parameter is small (0.9/11.6 =
0.08), a PI controller should perform well for this loop. Using Table 4-1, the
QDR tuning parameters are:
Figure 4-2 shows the responses of the transmitter output PV and the con-
troller output OP to a step change in the disturbance variable. As expected,
the shorter integral time recommended by the QDR formula in comparison
with the IMC rule results in a faster return to the set point with about the
same oscillatory behavior. This is at the expense of a slightly larger initial
change in the controller output.
How to Tune Feedback Controllers 63
Figure 4-2. Responses of PI Controllers for a Very Controllable Loop Tuned with
QDR and IMC Formulas
TI = 3 min
TI = W = 11.6 min
KKc = 11.6
TI = 3 min
TI = W = 11.6 min
An open-loop test on a loop results in a time constant of 3.5 min and a dead
time of 4.5 min. For this low-controllability loop, a PID controller is indi-
cated. The QDR tuning parameters from Table 4-1 are:
Figure 4-3 shows the responses of the transmitter output PV and the con-
troller output OP to a step change in the disturbance variable. As expected
the shorter integral time proposed by IMC returns the PV to the set point
faster, and in this case without oscillation. Besides, this is accomplished
here without the expense of a higher overshoot in the initial change in con-
troller output.
64 Tuning of Industrial Control Systems, Third Edition
Figure 4-3. Responses of PID Controllers For Low Controllability Loop Tuned
with QDR and IMC Formulas
TI = W = 3.5 min
TI = 9 min
KKc = 0.93
TD = 2.2 min
TI = W = 3.5 min
TI = 9 min
In the preceding examples, the responses are to a step change in the distur-
bance variable, such as that of the process flow to an exchanger. In continuous
processes, the objective of the control system is to maintain the process vari-
able at or near the set point in the presence of disturbances. The set point is
seldom changed. However, what happens when there is a need for the opera-
tor to change the set point? When the controller gain is high, a sudden change
How to Tune Feedback Controllers 65
in set point can cause a large change in the controller output, as shown in Fig-
ure 4-4. The response shown in Figure 4-4 is to a 3F change in set point for the
process of Example 4-2. Since the loop gain is 11.6, the figure shows that the
controller output initially changes by over 30%. Note that this is for a process
gain K = 1; if the process gain is higher the change in output will be smaller,
but if the process gain is lower than usual the controller output will have a
larger change. This sudden large change in controller output is bound to cause
a disturbance to other loops. For example, if the controller manipulates the
steam flow as in Figure 3-1, there may be a drop in the steam header pressure
that will affect other systems on that steam line.
Figure 4-4. Responses of a Loop with High Proportional Gain to a Step Change in
Set Point (Set Point Shown as a Dashed Line)
SP
KKc = 11.6
TI = 3 min
66 Tuning of Industrial Control Systems, Third Edition
There are several ways to prevent sudden large changes in controller output
when the set point is changed:
2. Configure the controller so that the proportional mode acts on the process
variable and not on the deviation from the set point. When the set point is
changed, the output will ramp to the new value at a rate controlled by the
integral time.
4. Have the operator slowly change the set point in small steps.
Keep in mind that a large change in controller output happens only when the
controller gain is high, as in Example 4-2. This is not a problem with less con-
trollable loops such as the one in Example 4-3.
A similar problem could occur with the derivative mode except that the
default configuration for the derivative mode is to act on the process variable
and not on the deviation from set point.
There are cases in which set point changes are common, such as with batch
processes and on-line optimization. One recent development in industrial
operations is the incorporation of on-line optimization programs that auto-
matically change controller set points as the optimum conditions change.
Most of these programs contain limits on the size of the set point changes they
make. At any rate, as mentioned above, one sure way to prevent large changes
in controller output with set point changes is to have the proportional mode
act on the process variable instead of on the deviation from set point. As long
as there is integral mode, this option will not affect the performance of the
controller on the disturbance variables.
The examples in this section have illustrated the performance of the controller
when tuned with the parameters of the open-loop test. To summarize our
findings:
Except for controllers that must constantly respond to set point changes
(e.g., slaves in cascade loops; see Chapter 7), the controller should be
tuned for good performance on disturbance inputs, and sudden set
point changes should be limited in magnitude.
For very controllable processes, the tuning formulas call for high loop
gains.
The following is a collection of tips the authors hope will be useful in making
the controller-tuning task more efficient and satisfying.
When the formulas result in higher gains and shorter integral times than
seem reasonable, let your judgment override the formulas.
All of these problems cause poor feedback controller performance that must
be handled by means other than controller tuning; for example decoupling,
feedforward control, adaptive control, or the use of valve positioners. The
chapters that follow present these techniques.
70 Tuning of Industrial Control Systems, Third Edition
The problem of reset windup or saturation of the controller output is one that
may often be considered a tuning problem, when in reality it cannot be
resolved by tuning the controller. It is therefore important to be able to recog-
nize the symptoms of reset windup and to know how to resolve them.
A properly tuned controller will behave well as long as its output remains in a
range where it can change the manipulated variable, but it will behave poorly
if, for some reason or other, the effect of the controller output on the manipu-
lated flow is lost. A gap between the limit on the controller output and the
operational limit of the control valve is the most common cause of reset
windup. The symptom is a large overshoot of the controlled variable while
the integral mode in the controller is crossing the gap. Reset windup occurs
most commonly during startup and shutdown, but it can also occur during
product grade switches and large disturbances during continuous operation.
Momentary loss of a pump may also cause reset windup.
One way to prevent the overshoot of Figure 4-5 would be to start up with the
controller in Manual state. The console operator must then watch the temper-
ature and as it approaches the set point, set the controller output to, for exam-
How to Tune Feedback Controllers 71
Valve opened
ple, 50% or if known, to a value near the required output (20% in this
example) before switching the controller to Automatic.
The problem of reset windup can also occur during normal operation when
there is a gap between the limits of the controller output and the operating
limits of the valve position or other manipulated variable. For example, if in
case of Figure 4-5 the controller output were limited to -10% to 110% while the
valve operates between 0 and 100%, windup would occur when a large distur-
bance causes the controller output to enter the gap. This is because the con-
troller has no effect on the process variable while its output is in the gap and
the integral mode keeps the controller output in the gap until the process vari-
able crosses the set point, resulting in an overshoot.
Some processes exhibit what is known as inverse response; that is, an initial
move in a direction opposite to the final steady-state change when the input is
72 Tuning of Industrial Control Systems, Third Edition
Another typical example of inverse response is the level in the steam drum of
a water tube boiler when the steam demand changes. The inverse response is
caused by the phenomena of swell and shrink of the steam bubbles in the
boiler tubes.
When the inverse response is to a change in the controller output, the loop
becomes very uncontrollable. For example, if in the reactor of Figure 4-6 the
temperature were controlled by manipulating the reactant flow, every action
by the controller would be followed by an immediate change in the tempera-
ture in the wrong direction. This would be worse than if the process had a
dead time equal to the duration of the inverse response. Fortunately, such a
situation is extremely rare, but it should be kept in mind when troubleshoot-
ing difficult tuning problems.
74 Tuning of Industrial Control Systems, Third Edition
As discussed in Section 3-5, most processes exhibit nonlinear behavior; that is,
their sensitivity to changes in controller output and dynamic behavior change
with operating conditions. This means that although the controller can per-
form well at the set of operating conditions at which it is tuned, the perfor-
mance can deteriorate at some other operating conditions. One characteristic
of many processes is that they become more sensitive at lower throughput
rates.
Figure 4-7 shows the temperature response of the process heater of Figure 3-1
at full production rate and at half production rate. The controller is PID tuned
for QDR response at full production rate. The figure shows that at half
production rate, the response becomes highly oscillatory. This is because there
is half as much process fluid to absorb the heat provided by the steam, making
the temperature twice as sensitive to the action of the controller. As discussed
in Section 3-5, an equal-percentage control valve will provide a lower loop
gain at lower controller outputs to compensate for the higher process
sensitivity.
4-6. Summary
The effectiveness of the tuning formulas was demonstrated for normal pro-
cesses as well as for very controllable and very uncontrollable processes. The
advantage of the PID over the PI controller was also demonstrated.
Helpful tuning hints were presented for those instances when the open-loop
test cannot be performed.
Figure 4-7. The Response Is More Oscillatory at Half Production Rate because the
Loop Gain is Twice as High than at Full Rate
earity. The problem of large initial changes in the controller output when the
set point is changed and when the controller gain is high was also discussed.
The next chapter presents the selection of controller modes and tuning for a
number of common control loops.
References
Review Questions
4-1. Based on the tuning formulas given in this chapter, how must you
change the controller gain if, after the controller is tuned, the process
gain were to double because of the nonlinear behavior of the process?
4-3. Assuming that the quarter-decay-ratio formulas of Table 4-1 give the
same tuning parameters as those of Table 2-1, what relationship can be
established between the controller ultimate gain and the gain with the
uncontrollability parameter of the process in the loop? What is the rela-
tionship between the ultimate period and the process dead time?
4-5. Estimate the tuning parameters of a PID controller for the three pro-
cesses of question 4-4.
4-6. Why would one want to configure the controller so that the proportional
mode acts on the process variable and not the deviation from set point?
What would the response of the controller output be when the controller
is configured as such and the set point is changed?
4-7. What is the typical symptom of reset windup? What causes it? How can
it be prevented?
4-8. What is known as inverse response? What effect does it have on the perfor-
mance of a feedback controller and why?
5
Mode Selection and
Tuning of Common
Feedback Loops
The preceding chapters dealt with the tuning of feedback controllers for gen-
eral processes that can be represented by a single-lag-plus-deadtime
(SLPDT) model. This chapter presents tuning guidelines for the most typical
process control loops, specifically flow, level, pressure, temperature, and com-
position control loops.
C. Design and tune simple feedback controllers for flow, level, pressure,
temperature, and composition.
77
78 Tuning of Industrial Control Systems, Third Edition
Although the most common objective for feedback control is to maintain the
controlled variable at its set point, there are some control situations, often in
the control of level or pressure, when it is acceptable to just maintain the con-
trolled variable within an acceptable range. The difference between these two
objectives is important because, as Chapter 2 showed, the purpose of the inte-
gral mode is to eliminate the offset or steady-state deviation of the process
variable from the set point. Consequently, integral mode is not required when
it is acceptable to allow the controlled variable to vary over a range. One
advantage of eliminating the integral mode is that it permits higher propor-
tional gain, thus reducing the initial deviation of the controlled variable
caused by disturbances.
The second situation calls for proportional controllers with as wide a propor-
tional band as possible. These are found in the control of level in intermediate
storage tanks and condenser accumulators, and in the control of pressure in
Mode Selection and Tuning of Common Feedback Loops 79
gas surge tanks, because in these cases the purpose of the tank is to attenuate
variations in process flow.
Flow control is the simplest and most common of the feedback control loops.
The schematic diagram of a flow control loop in Figure 5-1 shows that there
are no lags between the control valve that causes the flow to change and the
flow sensor/transmitter (FT) that measures the flow. Since most types of flow
sensors (orifice, venturi, flow tubes, magnetic flowmeters, turbine meters,
coriolis, etc.) respond very fast, the only significant lag in the flow loop is the
control valve actuator, and most actuators have time constants of the order of
a few seconds.
SP
FC
FT
However, when the flow controller is the slave in a cascade control scheme
(see Chapter 7), it is important for the flow to respond quickly to set point
changes. This requires a proportional-integral controller with a gain near
unity, which to maintain stability may require an increase in the integral time
from the few seconds normally used in flow controllers. The IMC2 tuning
rules (see Section 4-1) suggest that the integral time be set equal to the time
constant of the loop, usually that of the control valve actuator. In cascade situ-
ations, tight flow control is indicated.
The proportional gain should also be increased when hysteresis of the control
valve causes variations in the flow around its set point. As mentioned in Sec-
tion 4-2, hysteresis is caused by static friction in the valve packing that creates
a difference between the actual valve position and the corresponding control-
ler output. The error changes direction according to the direction in which the
valve stem must move, and this causes a dead band around the desired valve
position; that is, a band within which the valve does not respond to changes in
the controller output. Increasing the flow controller gain reduces the ampli-
tude of the flow variations caused by hysteresis. A valve positioner also
reduces hysteresis and speeds up the valve response, but positioners are usu-
ally difficult to cost-justify for flow control loops.
Figure 5-2 shows responses of a flow control loop with valve hysteresis for
two different tunings of the controller. The top curve is for the traditional
tuning of low gain and fast integral (that is, a short integral time), while the
bottom curve is for a more aggressive tuning of a gain of 1.5 and the same
integral time. As the figure shows, the more aggressive tuning reduces the
variations in flow caused by hysteresis in the valve.
Figure 5-2. Oscillations on a Flow Control Loop with Valve Hysteresis are
Reduced in Amplitude with a Higher Controller Gain
Keeping level and pressure constant calls for tight control, while smoothing
out variations in control usually calls for averaging control. Pressure is to
gas systems what level is to most liquid systems, although liquid pressure is
sometimes controlled.
Tight Control
One example of tight liquid level control and one example of tight pressure
control are shown in Figure 5-3. The control of level in natural-circulation
evaporators and reboilers is important because too low a level causes deposits
on the bare hot tubes and overheating of the tubes at the top. Conversely, too
high a level causes elevation of the boiling point, reducing the heat transfer
rate and preventing the formation of bubbles, which enhances heat transfer by
promoting turbulence. The example of tight pressure control or pressure regu-
82 Tuning of Industrial Control Systems, Third Edition
The design of tight level and pressure control systems requires a fast-acting
control valve, with a positioner if necessary, to prevent secondary time lags
that would cause oscillatory behavior at high controller gains. If the level or
pressure controller is cascaded to a flow controller, the latter must be tuned as
tight as possible, as discussed in the preceding section.
Normally, only proportional mode is needed for tight level or pressure con-
trol. The proportional gain must be set high, from 10 to over 100 (proportional
band of 1 to 10% of range). If the lag of the level or pressure sensor is signifi-
cant, derivative mode could be added to compensate for it and to afford a
higher gain. The derivative time should be set approximately equal to the time
constant of the sensor (see the next section). Integral mode should not be used,
since it would require a reduction of the proportional gain.
Two examples of averaging level control are shown in Figure 5-4: the control
of level in a surge tank (a) and in a condenser accumulator drum (b). Both the
surge tank and the accumulator drum are intermediate process storage tanks.
The liquid level in these tanks has absolutely no effect on the operation of the
process. It is important to realize that the purpose of an averaging level con-
troller is to smooth out flow variations while keeping the tank from overflow-
ing or running empty. If the level were to be controlled tightly in such a
situation, the outlet flow would vary just as much as the inlet flow(s), and it
would be as if the tank (or accumulator) were not there.
Figure 5-3. Examples of Tight Control: (a) Evaporator Level; (b) Supply Header
Pressure
Vapors
Feed
Steam
LC LT T
Condensate
(a)
Product Supply
PC
PT
(b) Loads
84 Tuning of Industrial Control Systems, Third Edition
Figure 5-4. Examples of Averaging Level Control: (a) Surge Tank; (b) Condenser
Accumulator
Feeds
SP
Surge
tank LT LC
Outlet flow
(a)
Vapors
Condenser
Accumulator LT LC
Column
FC
FT Distillate
(b)
Reflux
Mode Selection and Tuning of Common Feedback Loops 85
at the set point, a term that is only important when the controller has no inte-
gral mode. This will cause the outlet valve to be fully open when the level is at
100% of range and fully closed when the level is at 0% of range, thus using the
full capacity of the valve and of the tank. A proportional gain higher than
unity would reduce the effective capacity of the tank for smoothing variations
in flow, while a gain lower than unity would reduce the effective capacity of
the control valve and create the possibility of the tank overflowing or running
dry. With the proposed design the tank behaves as a low-pass filter to flow
variations; a low-pass filter allows low-frequency input through while it
attenuates high frequency variations. The time constant of such a filter is:
A ( h max h min )
= -----------------------------------------
- (5-1)
K c F max
where:
hmin and hmax = the low and high points of the range of the level
transmitter, respectively, ft
Fmax = the maximum flow through the control valve when opened fully
(100% controller output), ft3/min
The controller gain is assumed to be 1.0 in this design. When the level control-
ler is cascaded to a flow controller, Fmax is the upper limit of the range of the
flow transmitter in the flow control loop. Note that a proportional gain greater
than unity results in a reduction of the filter time constant and therefore less
smoothing of the variations in flow. A good way to see it is to note that dou-
bling the gain would be equivalent to reducing either the tank area or the
transmitter range by a factor of two, thus reducing the effective capacity of the
tank. On the other hand, reducing the controller gain to half would be equiva-
lent to reducing the capacity of the valve by half, thus increasing the possibil-
ity of the tank overflowing.
their set points. The process in a level control loop is unlike most other loops
in that it does not self-regulate; that is, the level tends to continuously rise or
fall when the feedback controller is not in Automatic. This usually means that
for level control loops, a time constant cannot be determined by an open-loop
test. Even when there is some degree of self-regulation, the process time con-
stant is very long, on the order of one hour or longer. Because of this, PI con-
trollers in level control loops have the following characteristics:
The level, and the flow that is manipulated to control it, oscillate for a
long period. Sometimes the period is so long that the oscillation is
imperceptible, unless it is trended over a very long time.
The shorter the integral time, the shorter the period of oscillation.
The level control loop is unstable when the integral time is equal to or
shorter than the time constant of the control valve.
Unlike most other loops, there is a range of controller gains over which
the oscillations increase as the controller gain is decreased.
This leads to the following general rules for tuning PI controllers for averag-
ing level control:
There are intermediate situations that do not require a very tight level control
but where it is important not to allow the level to swing over the full range of
the transmitter, as in averaging level control. A typical example is a blending
tank, where the level controls the tank volume, and therefore the residence
time for blending. If a 5% variation in residence time is acceptable, a propor-
tional controller with a gain of 5 to 10 or even lower could be used, since the
flow would not be expected to vary over the full range of the control valve
capacity.
Mode Selection and Tuning of Common Feedback Loops 87
Figure 5-5 shows responses of the control of the level in a tank such as the
one in Figure 5-4(a) with the level controller tuned for averaging and for
tight level control. The inlet flow into the tank increases by increments of
200 gpm as several batch processes dump their contents into the tank. The
tank has a total capacity of 10,000 gallons, while the valve has a flow capac-
ity of 1,000 gpm when fully open.
Figure 5-5 also illustrates that averaging level control (the continuous lines)
averages out the variation of the inlet flow, resulting in a smooth variation
of the outlet flow. On the other hand, tight level control (the dashed lines)
maintains the level nearly constant, but this requires that the outlet flow
essentially follow the variations in the inlet flow, just as if the tank were
not there. In this example the averaging level controller has a gain of 1.0 and
the tight controller has a gain of 20; both have integral times of 20 minutes.
MC
s = ------------p- (5-2)
hA
where:
Figure 5-5. Responses of Averaging Level Control (Continuous Lines) and Tight
Level Control (Dashed Lines) on a Surge Tank
Level
Kc = 1
Kc = 20
TI = 20 min
Outlet flow
Kc = 20
Kc = 1
When these units are used, the time constant is calculated in seconds.
in the exchanger, which is proportional to the flow and to the change in tem-
perature of the hot oil:
Q = FoilCp(Toin - Toout)
where:
This calculation is carried out in the heat rate controller QC in Figure 5-7 to
determine the process variable of the controller. The process outlet tempera-
ture controller TC sets the set point of the heat rate controller QC.
SP
Process TC
flow
TT
Air Fuel
Hot oil
Toin TT SP
TC
Foil FT
SP
QC
Process TT
in
Process
out
Toout TT
In spite of all the sources for time delays in sampling and analysis, since it is
the ratio of the dead time to the process time constant that determines the
uncontrollability of the loop (see Chapter 4), if the combination of the analysis
sampling period and time delay is less than the process time constant, a pro-
portional-integral-derivative (PID) controller is indicated. The tuning strategy
of Chapters 2 and 4 can be used. On the other hand, if the total dead time is on
92 Tuning of Industrial Control Systems, Third Edition
the order of several process time constants, theories such as IMC2 and control-
ler synthesis1 call for a pure integral controller. This is because the process
responds fast relative to the time frame in which the analysis is done. Chapter
6 discusses the tuning of controllers that make use of sampled, rather than
continuous, measurements.
5-6. Summary
This chapter presented some guidelines for selecting and tuning feedback
controllers for several common process variables. While flow control calls for
fast PI controllers with low gains, level and pressure control can be achieved
with simple proportional controllers with high or low gains, depending on
whether the objective is tight control or smoothing of flow disturbances.
When PI controllers are used for level control, the integral time should be
long, on the order of one hour or longer. PID controllers are commonly used
for temperature and analyzer control.
References
Review Questions
5-1. Briefly state the difference between tight level control and averaging
level control. In which of the two is it important to maintain the level at
the set point? Give an example of each.
5-2. What type of controller is recommended for flow control loops? Indicate
typical values for the gain and integral times.
5-3. What type of controller is indicated for tight level control? Indicate typi-
cal gains for the controller.
Mode Selection and Tuning of Common Feedback Loops 93
5-4. What type of controller is indicated for averaging level control? Indicate
typical gains for the controller.
5-5. When a PI controller is used for averaging level control, what should the
integral time be? Would an increase in gain increase oscillations or
decrease oscillations?
5-6. Estimate the time constant of a temperature sensor weighing 0.03 kg,
with a specific heat of 23 kJ/kg-C. The thermowell has a contact area of
0.012 m2 and the heat transfer coefficient is 0.6 kW/m2-C.
5-7. Why are PID controllers commonly used for controlling temperature?
This chapter deals with tuning methods for loopssuch as analyzer control
loopsin which the process variable cannot be measured continuously. In
such loops the process variable must be sampled at discrete intervals of time,
at which the control calculations are carried out and the controller output is
updated, to be kept constant until the next update.
Learning ObjectivesWhen you have completed this chapter you should be able
to:
95
96 Tuning of Industrial Control Systems, Third Edition
When the process variable is measured continuously, the control system sam-
ples it fast enough relative to the response time of the process for the sampling
to be of no consequence to the performance of the controller. In such cases, the
algorithm used to do the PID calculations is as transparent to the control engi-
neer as the pneumatic and electronic circuits that are used to determine the
controller output in analog controllers. However, when the sampling interval
is of the order of the process response time, the control algorithm becomes of
interest.
Since the feedback control calculation is made at regular intervals of time, the
process variable (PV) is sampled only when the controller output (OP) is cal-
culated and updated. The controller output is updated at the sampling
instants and is held constant for one sampling interval (the period of time
between output updates) of duration T.
Formulas for carrying out the PID algorithm calculations are given in Table 6-
1. In this algorithm, the process variable (PVk) is first used to calculate Yk, the
output of a proportional-derivative (PD) calculation. This is done to avoid
undesirable pulses in the controller output on set point changes by having the
Tuning Sampled-Data Control Loops 97
derivative mode work on the process variable PVk instead of on the deviation
from set point. The PD calculation also contains a filter with time constant
TD, which is intended to limit the magnitude of pulses in the controller out-
put upon sudden changes in the process variable. It is seldom desirable for the
derivative mode of the controller to respond to set point changes, because on a
set point change there would be a large change in the controller output lasting
for just one samplethat is, a large undesirable output pulse known as a
derivative kick. Such pulses are completely avoided by the algorithm of
Table 6-1 since the derivative mode, acting on the process variable, does not
see changes in set point.
T D T ( + 1 )T D
Y k = -------------------- Y k 1 + -------------------- PV k + ------------------------- ( PV k PV k 1 )
T + T D T + T D T + T D
Ek = SPk - Yk
T
M k = K c E k E k 1 + ------ E k
T I
Controller output:
Mk = Mk-1 + Mk
where:
The deviation from the set point or error (E) in Table 6-1 is for a reverse acting
controller. For a direct-acting controller the terms Yk and SPk are reversed in
the formula or, alternatively, the controller gain is set to a negative value.
When either the derivative time TD is set to zero (PI controller) or the process
98 Tuning of Industrial Control Systems, Third Edition
variable reaches a steady value, Yk = PVk, the algorithm still drives the pro-
cess variable to the set point.
The filter parameter of Table 6-1 has a special meaning; its reciprocal, 1/, is
the amplification factor on the change of the PV at each sampling instant and
is also called the dynamic gain limit. Note that, if were set to zero, the
amplification factor on the change in PV would have no limit. For example, if
the sampling interval is one second (1/60 min) and the derivative time is one
minute, the change in PV at each sample with = 0 would be multiplied by a
factor of 60 (TD/T = 60). By setting the nonadjustable parameter to a
reasonable value, say 0.1, the change in PV cannot be amplified by a factor
greater than 10, independent of the sampling interval and the derivative time.
The dynamic limit permits setting the derivative time to any desired value
without the danger of introducing large undesirable pulses in the controller
output.
Calculate the output of the derivative term on the PD unit (the equation that
calculates Yk in Table 6-10) of the PID control algorithm to a ramp that
starts at zero and increases by 1% with each sample. Use a sampling interval
of 1 s and a derivative time of 0.5 min. The derivative filter parameter is =
0.1.
Substitution of the values given and of the process variable at each sample
into the series controller of Table 6-1 produces the results summarized in
Table 6-2. The results for the ideal derivative unit are calculated using the
filter parameter = 0.
For example, for the value of Y1, at the first sample Y0 = 0, PV1 = 1%,
PV0 = 0%, TD = 0.5(60) = 30 s, so:
Notice that the unfiltered (ideal) derivative unit jumps to 30 at time 0 with
increments of 1 at each sample. Both of these responses are shown graphi-
cally in Figure 6-1. The unfiltered derivative unit is leading the PV by one
derivative time (30 s), while the derivative unit with the filter, after a brief
lag, also leads the PV by one derivative time. In practice, the lag is too small
to significantly affect the performance of the controller.
80
70
60
Yk
50
40
D=0
30
PVk
20 D = 0.1
10
TD = 30 T
0
0 5 10 15 20 25 30 35 40 45
t/T
100 Tuning of Industrial Control Systems, Third Edition
Similar to the derivative kick, the sudden change in controller output caused
by the proportional mode right after a change in set point is known as pro-
portional kick, although it is not a pulse. It too can be eliminated by replacing
the deviation Ek with the output of the derivative unit Yk in the calculation of
the increment in controller output in the control algorithm of Table 6-1. Mod-
ern computer- and microprocessor-based controllers offer the option of hav-
ing the proportional mode act on either the deviation or the process variable.
This option, known as proportional-on-PV, must be selected on the following
basis:
If the controller is the slave of a cascade control scheme (see Chapter 7),
the proportional mode must act on the deviation from set point. Other-
wise, when the main controller changes the set point of the slave, the
slave will not respond immediately, as it must for the cascade scheme
to work.
where:
The bars around the error indicate the absolute value or magnitude of the
error. By using the absolute value of the error, the gain increases when the
error increases in either the positive or the negative direction.
The nonlinear gain option is normally used with averaging level controllers
(see Section 5-3) because it allows a wider variation of the level near the set
point while still preventing the tank from overflowing or running dry, as
illustrated in Figure 6-2. In the figure, the gain at the middle of the range is
0.25 (the dashed line) and unity is at the two extremes of the range. The non-
linear gain allows greater smoothing of flow variations with a given tank; that
is, it makes the tank look as if it has a larger capacity than it does, as long as
the flow varies near the middle of its range. Some computer controllers pro-
vide the option of having a zero gain at zero error, a feature which is desirable
in some pH control schemes.
100%
80%
60%
OP
40%
20%
0%
0% 20% 40% 60% 80% 100%
PV
Prevention of the tanks overflowing or running dry requires that the valve
be fully open when the level is at 100% of the range and fully closed when
the level is at 0%. Since the set point is 50%, either requirement takes effect
when the magnitude of the error is 50%. With the output bias of 50%,
using the upper limit requirement in Equation 6-1:
Recall from Equation 5-1 that the time constant of the tank is inversely pro-
portional to the controller gain, thus the effective capacity of the tank, as
used for smoothing flow variations, can be increased from its real value at
full and zero flow to four times that value at one-half of full flow.
Section Summary
When the controller is tuned using the process parameters of gain, time con-
stant, and dead time, estimated as shown in Chapter 3, the effect of sampling
is not included in the process model. This is because the process model is
obtained from a step test in controller output, as you can recall from Chapter
104 Tuning of Industrial Control Systems, Third Edition
3, and such a step will always take place at a sampling instant and will remain
constant after that.
Moore and his coworkers3 developed a simple correction of the controller tun-
ing parameters for the effect of sampling. They point out that when a continu-
ous signal is sampled at regular intervals of time and then reconstructed by
holding the sampled values constant for each sampling period, the recon-
structed signal is effectively delayed by approximately one-half the sampling
interval, as shown by the dashed line in Figure 6-3. In sampled-data systems
the controller output is held constant between updates, thus adding one-half
the sampling interval to the dead time of the rest of the loop components. The
correction for sampling is then simply to add one-half the sampling interval to
the dead time obtained from the step response. The uncontrollability parame-
ter is then given by:
T
t 0 + ---
2
P u = --------------- (6-2)
where:
The formulas of Table 6-3 contain an adjustable parameter q that affects only
the controller gain. This parameter is adjusted in the range of 0 to 1 to shape
the tightness of the closed-loop response. If the model parameters were an
Tuning Sampled-Data Control Loops 105
Figure 6-3. Effective Delay of Sampling and Holding a Signal is One Half the
Sampling Interval
K = gain
= time constant, min
t0 = dead time, min
T = sampling interval, min
q = an adjustable parameter, in the range of 0 to 1.
t T
---
Let N = -----o- a = e
T
( 1 q )a
Proportional gain: K c = -------------------------------------------------------------------------------
K( 1 a)[1 + N(1 q)]
aT
Integral time: T I = ----------------
1a
106 Tuning of Industrial Control Systems, Third Edition
exact fit to the process response, the value of q would be the fraction of the
error at any one sample that would remain after one dead time plus one sam-
pling interval. For example, setting q = 0 specifies the process variable to
match the set point after N + 1 samples, where N is the number of samples of
dead timethat is, the sample time divided by the sampling interval. This
would result in the highest gain, and therefore in the tightest control. How-
ever, for any value of q, the tightness of the closed-loop response depends on
the ratio of the sampling interval to the time constant, T/. A more fundamen-
tal adjustable parameter is the closed-loop time constant c, which can be
related to the time parameters of the loopshort for fast processes and long
for slow processes. If c is specified, the value of q can be computed by:
T-
---
c
q = e
Setting q = 0 results in an upper limit for the controller gain. This value can be
used as a guide for the initial tuning of the controller. As is the case with the
tuning formulas of Chapter 4, the upper limit of the controller gain decreases
with increasing the process dead time, parameter N, in number of samples.
The formulas of Table 6-3 are intended to tune only a PI controller. Two time
constants plus a dead time would be required to tune a PID controller by this
procedure, but it is difficult to accurately determine more than one time con-
stant and a dead time from a simple open-loop step test.
As mentioned earlier, the formulas of Table 6-3 are applicable to any value of
the loop parameters and the sampling interval; moreover, the controller gain
can be adjusted to obtain fast response with reasonable variation of the con-
troller output. They are highly recommended because they relate the integral
and derivative times to the process time constants, thus reducing the tuning
procedure to the adjustment of the controller gain.
Use the tuning formulas of Table 6-3 to tune the temperature controller for
the heater of Section 3-1. Use a PI controller with sampling intervals of 1 s,
10 s, 30 s, 1 min, 2 min, and 4 min.
Tuning Sampled-Data Control Loops 107
The process parameters for the heater were determined in Section 3-1:
The tuning parameters, for a sampling interval of 0.5 min and q = 0, are:
0.5
-------
7.5
N = 2.5 0.5 = 5 a = e = 0.936
TD = 0
For the other sampling times the tuning parameters are given in Table 6-4.
Notice that the gain is lower and the integral time is shorter as the sampling
interval is increased. This means that the loop is less controllable at the longer
sampling interval. Also notice that there is a very small change in the tuning
parameters when the sampling frequency is increased from 6 times per min-
ute to 60 times per minute. Since most control systems sample at the rate of
more than once per second, the effect of sampling is negligible for most con-
trol loops.
Figure 6-4 shows the responses of the heater temperature controller to a step
change in disturbance with the tuning parameters given in Table 6-4 and sam-
pling intervals of 0.5, 1.0, and 2.0 min. The figure shows that there is little dif-
ference in controller performance for sampling interval times of 0.5 and 1 min,
but the performance deteriorates with a sampling interval of 2 min.
108 Tuning of Industrial Control Systems, Third Edition
T = 1 min
T = 0.5 min
T = 2 min
When the sampling interval is more than three or four times the dominant
process time constant, the process reaches steady state in responding to each
controller output move before it is sampled again. This may happen because
the process is very fast or because the sensor is an analyzer with a long cycle
time. For such situations the formulas of Table 6-2 result in a pure integral
controller:
where:
1q
K 1 = -----------------------------------------
K[1 + N(1 q)]
Tuning Sampled-Data Control Loops 109
Notice that for the case N = 0 and q = 0, the controller gain is the reciprocal of
the process gain. This result makes sense, since a loop gain of 1.0 is what is
needed to reduce the error to zero in one sampling interval if the process
reaches steady state during that interval. An interesting application of this
case is a chromatographic analyzer sampling a fast process. The nature of
such an analyzer is that the composition is not available to the controller until
the end of the analysis cycle, because it takes a full analyzer cycle to separate
the mixture and analyze it. This means that the process dead time is approxi-
mately one sampling interval, or N = 1. For q = 0, Equation 6-3 gives a gain of
KI = 1/K(1 + 1) = 1/2K, or one-half the reciprocal of the process gain. This also
makes sense, because when action is taken by the controller, it takes two sam-
pling intervals for the controller to see the result of that action, so the formula
says to spread the corrective action equally over two samples.
For the process heater of Section 3-1, calculate the maximum gain for the PI
controller using the formulas of Table 6-2 and sampling times of 5, 10, and
20 min. Also calculate the gain of the pure integral controller, given by KcT/
TI (this is the same as the KI of Equation 6-3).
ters are those shown in Table 6-5 except that the gain was reduced by half (q
= 0.5) with the sampling interval of 5 min, and to one-fourth (q = 0.75) with
the sampling intervals of 10 and 20 min. As expected, controller perfor-
mance is a lot slower than it is in the responses of Figure 6-4, but still
acceptable when the sampling interval is forced to be that long. This shows
that the tuning formulas of Table 6-2 can be applied to a wide range of sam-
pling-interval-to-time-constant ratios.
T = 5 min
q = 0.5
T = 10 min
q = 0.75 T = 20 min
q = 0.75
To summarize, the formulas presented in Table 6-2 can be used with the open-
loop test model, resulting in a PI controller. They are applicable over a wide
range of sampling intervals and dead-time-to-time-constant ratios. Although
formulas used to tune PID controllers have not been developed by this
method, this does not present a problem because the derivative mode should
not be effective when the process variable is sampled slowly.
Tuning Sampled-Data Control Loops 111
It makes sense to ratio the sampling interval to the process time constant
because the relative change in the process output from one sample to the next
depends only on this ratio; that is, the relative change will be the same for a
process with a 1-minute time constant sampled once every 5 seconds as for a
process with a 10-minute time constant sampled every 50 seconds.
What is the problem with too long a sampling interval? The problem is that
for a very slow process the increments in controller output at each update
become very small and may be lost in the precision with which the controller
output is calculated. The reason the controller output becomes small is
because:
112 Tuning of Industrial Control Systems, Third Edition
For example, the pressure control loop of an ammonia synthesis loop, where
the pressure is controlled with the purge flow, has a time constant of about 40
minutes (2,400 s). If the controller is updated 10 times per second and the inte-
gral time is of the order of magnitude of the time constant, the term T/TI in
Table 6-1 is 0.1/2,400 = 0.00004. This makes the controller output, when the
error is 1% and the output 50%, equal to 50.00004%, so that there is an imper-
ceptible change on the position of the valve after each update of the controller
output. The reason this is not commonly a problem is that the controller out-
put is usually computed with double precision (twice as many digits as nor-
mal in the representation of the output result) and the output can be
accurately computed so that after many samples (about 10,000 in this case),
the small increments are not lost.
In our experience, the control loop performance is acceptable when the con-
troller update frequency is as high as one-tenth the loop time constant. At one
time, Fisher Controls marketed a very successful computer control package
for ammonia plants that used a sampling interval of 5 minutes to control the
synthesis loop pressure with the purge flow.
It has been clearly established in Chapter 4 and Section 6-1 that feedback
controllers cannot perform well when the process has a high ratio of dead
time to time constant. The total loop gain must be low for such processes, so
the deviations of the controlled variable from its set point cannot be kept low
in the presence of disturbances. One way to improve the performance of the
feedback controller for low controllability loops is to design a controller that
compensates specifically for the process dead time. This section presents two
controllers that have been proposed to compensate for dead time: the Smith
Predictor and the Dahlin Dead Time Compensation Controller (Dahlin
Controller).
Dead time compensation requires that past values of the controller output be
stored and played back. Not until the advent of computer- and microproces-
sor-based controllers was the storage and playback of control signals possible.
The computer memory provides the ability to store and retrieve past sampled
values.
ler, so that it ends up with too many adjustable parameters: the model param-
eters plus the controller tuning parameters. Because there are so many
parameters to adjust, there is no convenient way to adjust the closed-loop
response when the model does not properly fit the process. Given the nonlin-
ear nature of process dynamics, any technique that depends heavily on exact
process modeling is doomed to fail.
Figure 6-6. Block Diagram of the Smith Predictor for Dead Time Compensation
Disturbances
SP OP Process PV
Controller with Dead
+ Time
-
+
Process Model -
Model Dead Time
+
Corrected model output + Model error
The Dahlin dead time compensation controller can be reduced to a PID con-
troller with an extra term. The only modification to the controllers of Table 6-1
is in the calculation of the controller output:
where Mk can be computed by the control algorithm of Table 6-1. The last
term in the calculation of the output provides the dead time compensation.
Note that the term vanishes when there is no dead time, N = 0. The actual con-
troller is tuned with the formulas of Table 6-2 except for the controller gain,
which is given by:
( 1 q )a
K c = --------------------- (6-5)
K(1 a)
The Dahlin Controller is used extensively to control processes with long dead
times. A common application is the control of paper machines, where the
properties of the paper can only be measured after it has gone through the
drying process, which introduces significant dead time. One characteristic of
this application is that the dead time is relatively constant and can be deter-
mined precisely. Dead time compensation presents problems in other pro-
cesses in which the dead time depends on flow and other process variables
(see Section 3-4).
Compare the response of the temperature controller for the process heater
of Figure 3-1 with and without dead time compensation. Use a PI controller
with a sampling interval of 1 min, which is approximately one-tenth of the
time constant (7.5 min). The open-loop parameters for the steam heater are a
gain of 1.95, a time constant of 7.5 min, and a dead time of 2.5 min.
The dead time compensation term requires two sampling intervals of dead
time:
N = int(t0/T) = int(2.5/1) = 2
116 Tuning of Industrial Control Systems, Third Edition
a = e(-1/7.5) = 0.875
Kc = (1-0.67)(0.875/0.125)/(1.95)[1+2(1-0)] = 0.4
Kc = (1-0.67)(0.875/0.125)/1.95 = 1.2
TI = 7.0 min
Figure 6-7 compares the steam heater responses of the controllers to a step
increase in process flow to the heater. The dead time compensation controller
results in a smaller deviation from set point and less oscillation than the regu-
lar PI controller. The improvement in performance is not spectacular, proba-
bly because this is a relatively controllable process. The value of q = 0.67 was
selected to prevent excessive oscillation in the controller output. The propor-
tional gains are then one-third, (1 0.67) = 0.33, of the maximum gains.
Figure 6-7. Response of Heater Temperature Control with and without Dead
Time Compensation
PI controller
with dead time
compensation
Standard PI
controller
6-5. Summary
References
Review Questions
6-3. How and why would you eliminate proportional kick on set point
changes? Will the process variable approach its set point faster or slower
when proportional kick is avoided? When must proportional kick be
allowed?
6-5. A process has a gain of 1.6%, a time constant of 20 min and a dead time
of 5 min. Calculate the tuning parameters for a discrete controller if the
sampling interval is (a) 4 s, (b) 1 min, (c) 5 min and (d) 50 min.
6-6. Repeat question 6-5, but for a controller with dead time compensation.
Specify also how many samples of dead time compensation, N, must be
used in each case.
6-7. What is the basic idea behind the Smith Predictor? What is its major dis-
advantage? How does the Dahlin Controller with dead time compensa-
tion overcome the disadvantage of the Smith Predictor?
7
Tuning Cascade Control
Systems
B. Select the control modes and tune the controllers in a cascade control
system.
121
122 Tuning of Industrial Control Systems, Third Edition
Figure 7-1 shows a typical cascade control system for controlling the tempera-
ture in a jacketed exothermic chemical reactor. (Note that the steam valve
shown in the figure is used at startup to preheat the reactor and is not relevant
to the following discussion.) The control objective is to control the tempera-
ture in the reactor, but instead of having reactor temperature controller TC 1
directly manipulate the jacket coolant valve, the jacket temperature is mea-
sured and controlled by a different controller, TC 2, which is the one that
manipulates the valve. The output of reactor temperature controller TC 1, the
primary controller, is connected or cascaded to the set point of jacket tem-
perature controller TC 2, the secondary controller. Notice that only the reac-
tor temperature set point is maintained at the operator-set value; the jacket
temperature set point varies to whatever value is required to maintain the
reactor temperature at its set point. A block diagram of the reactor cascade
control strategy, shown in Figure 7-2, clearly shows that the secondary control
loop is inside the primary control loop.
1. Any disturbances that affect the secondary variable are detected and com-
pensated by the secondary controller before they have time to affect the
primary control variable. Examples of such disturbances for the reactor of
Figure 7-1 are the coolant inlet temperature and pressure.
2. The controllability of the outside loop is improved because the inside loop
speeds up the response of the process dynamic elements between the con-
trol valve and the secondary variable. In the reactor example, the speed of
response of the jacket temperature is increased, resulting in a more con-
trollable loop for the reactor temperature.
3. Nonlinearities of the process in the inner loop are handled by that loop
and are removed from the more important outer loop. In the reactor exam-
ple, the nonlinear relationship between temperature and coolant flow is
made a part of the inner loop by the cascade arrangement, while the outer
loop enjoys the linear relationship between reactor and jacket tempera-
tures. Since the secondary loop should be more controllable than the pri-
mary loop, variations in the process gain are less likely to cause instability
when isolated in the secondary loop.
Tuning Cascade Control Systems 123
SP
TC
1
Reactants
TT
TC TT 1
2 2
Water
out
Coolant
Products
Steam
Disturbances
TJ TR
SP + Master + Slave
Controller Controller Jacket Reactor
TC-1 TC-2
- -
PV2 Jacket
TT
2
PV1 Reactor
TT
1
Disturbances into the inner loop will not be eliminated fast enough to
avoid their affecting the primary control variable.
Besides the inner loop having to be faster to respond than the outer loop, the
success of cascade control also requires that the sensor of the inner loop be fast
and reliable. One would not consider, for example, cascading a temperature
controller to a chromatographic analyzer controller. On the other hand, the
sensor for the inner loop does not have to be accurate, only repeatable,
because the integral mode in the primary controller compensates for errors in
the measurement of the secondary variable. In other words, it is acceptable for
the inner loop sensor to be wrong as long as it is consistently wrong, and to
the same degree.
level and gas pressure control loops, or when the controlled variable does not
have to be maintained tightly around its set point, as in averaging level con-
trol. When a level controller is cascaded to a flow controller, it is usually justi-
fied by the greater flexibility in the operation of the process, not by improved
control performance.
The above sections have examined the reasons and requirements for using
cascade control. The following sections will look at how to select the controller
modes for cascade control systems and how to tune them.
In a cascade control system, the primary controller has the same function as
the controller in a single feedback control loop: to maintain the primary pro-
cess variable at its set point. It follows that the selection of controller modes
for the primary controller should follow the same design guidelines presented
for a single controller in Chapter 5. On the other hand, the function of the sec-
ondary controller is not the same as that of the primary or single controller; it
therefore requires different design guidelines.
If the gain of the secondary controller is greater than unity, changes in the pri-
mary controller output result in higher immediate changes in the final control
126 Tuning of Industrial Control Systems, Third Edition
element than when a single feedback loop is used. This amplification results
in a faster response of the primary loop.
The use of integral and derivative modes in the secondary controller depends
on the application. Recall from previous chapters that adding integral mode
results in a reduction of the proportional gain, while adding derivative mode
results in an increase in the proportional gain. This may suggest that all sec-
ondary controllers should be proportional-derivative (PD) controllers, but this
is not the general case.
The integral mode should not be used in those secondary loops in which the
gain is limited by stability, and those in which the disturbances to the inner
loop do not cause large offsets in the secondary controller. The jacket temper-
ature controller of the reactor in Figure 7-1 is a typical example of a secondary
loop that does not require integral mode.
A common rule states that derivative mode should not be used in both the
secondary and primary controller, and since derivative would do the most
good on the least controllable loop, which is the primary loop, the rule essen-
tially reduces to never having derivative mode in the secondary controller.
This rule is based on the following reasoning:
1. Having all three modes in both the primary and secondary controller
results in requiring six tuning parameters, which without the proper
guidelines, makes the tuning task more difficult.
Tuning Cascade Control Systems 127
When the inner loop is fast and very controllable (e.g., flow loops), the sec-
ondary controller does not require derivative mode.
The tuning of the controllers in a cascade control system must be carried out
from the inside out; that is, the innermost loop must be tuned first, then the
loop around it, and so on. The block diagram of Figure 7-2 shows why this is
so: each inner loop is part of the process of the next outer loop.
Each loop in a cascade system must be tuned tighter and faster than the loop
around it; otherwise, the set point of the secondary loop would vary more
than its measured variable, resulting in poorer control of the primary process
variable. Ideally, the secondary variable should follow its set point as quickly
as possible, but with little overshoot and oscillations. If quarter-decay-ratio
(QDR) response is used for the secondary controller, the gain must be
adjusted to prevent excessive overshoot on set point changes. The ideal over-
shoot for the secondary process variable to a set point change is 5 to 10%.
128 Tuning of Industrial Control Systems, Third Edition
After each inner loop has been tuned in succession, the primary loop can be
tuned to follow any desired performance criteria by any of the methods of
Chapters 2, 4, 5 and 6. Given that what is special in cascade systems is the tun-
ing of the secondary loop(s), some typical secondary loops, namely flow, tem-
perature, and pressure, are briefly discussed next. Keep in mind, however,
that any variableincluding compositioncan be used as a secondary vari-
able provided it can be measured quickly and reliably (e.g., when a simple
continuous thermal conductivity detector is used to measure the hydrogen
composition in the ammonia synthesis process - see Example 7-2).
In modern computer control systems, flow is the innermost loop in most cas-
cade control schemes because it allows the operator to intervene in the control
scheme by taking direct control of the manipulated flow. Figure 7-3 shows a
typical temperature-to-flow control scheme. The flow transmitter compen-
sates for variations in the pressure drop across the control valve and absorbs
any nonlinearities of the valve. If the square root of the differential pressure is
extracted, the secondary measured variable and thus the output of the pri-
mary controller, becomes linear with the flow.
There are two difficulties with using temperature as the secondary measured
variable: the sensor lag and the possibility of reset windup. Fortunately, both
problems can be handled. The next section deals with the windup problem.
The sensor lag can be compensated for by using derivative mode in the sec-
ondary controller with the derivative time set equal to the sensor time con-
stant. As mentioned above, the derivative mode must act on the secondary
process variable only, not on the deviation from set point, to prevent having
two derivative units in series.
Tuning Cascade Control Systems 129
Vapors to
condenser
SP
TC
TT
Column
SP
FC
FT
Reflux
system. The pressure in the steam chest in the reboiler directly determines the
heat transfer rate because it controls the steam condensing temperature and
therefore the difference in temperature across the heat transfer area.
Reboiler
When both the primary and the secondary controllers are carried out by a
computer or microprocessor, the secondary (inner) loop is usually processed
Tuning Cascade Control Systems 131
at a higher frequency than the outer loop, so that the secondary controller has
time to respond to a set point change from the primary controller before the
next set point change takes place. Recall that the inner loop should respond
faster than the outer loop. If the sampling frequency is low and the same for
both the primary and secondary loops, the secondary loop must be processed
after the primary loop; otherwise, the change in set point will be delayed by
one sample before the secondary loop can take action.
This example shows how to tune the cascade control system for the jacketed
chemical reactor of Figure 7-1. For comparison, the response of a single reac-
tor temperature controller is compared to the response of the cascade con-
trol system. The single reactor temperature controller, TC 1, manipulates
the coolant valve directly, while in the cascade scheme the reactor tempera-
ture controller, TC 1, sets the set point of a jacket temperature controller, TC
2, which in turn manipulates the coolant valve, as in Figure 7-1. For the pur-
poses of this example, the manual steam valve is always closed.
To obtain the process parameters, a step test in the controller output con-
nected to the coolant valve is performed with the controllers in Manual, and
both the reactor temperature and the jacket temperature are recorded. The
following results are obtained:
Use the Ziegler-Nichols QDR tuning formulas of Table 4-1 to tune the sin-
gle reactor temperature PID controller:
Kc = 1.2(7.5/1.5)/2.2 = 2.8
The parameters from the response of the jacket temperature are used to
tune jacket temperature controller TC 2 in the cascade scheme. Since the
dead time is zero, a PI controller is indicated, the IMC rule of Section 4-1 is
used for the integral time and the gain can be as high as desired. To keep
the overshoot to set point changes reasonable:
Comparison with the results of the response to the step in coolant flow
shows that the reactor temperature loop has both a shorter time constant
and a shorter dead time when the jacket temperature controller is used.
Recall, however, that these parameters depend on the tuning of the jacket
temperature controller. For example, if a higher gain were used for TC 2, the
time parameters would be shorter still.
Tuning Cascade Control Systems 133
Kc = 1.2(4.5/0.75)/1.1 = 6.6
Figure 7-5 compares the responses of the single reactor temperature controller
and the cascade control scheme to a 5C step increase in coolant inlet tempera-
ture followed by a step decrease of 20% in the flow of the reactants. Since the
response of the outlet coolant temperature to the change in inlet coolant tem-
perature is immediately detected and corrected for by the secondary control-
ler, the reactor temperature in the cascade scheme hardly deviates from its set
point. The cascade scheme immediately increases the coolant flow to compen-
sate for the increase in inlet coolant temperature.
The figure shows that the cascade control scheme also improves the response
of the reactor temperature to a step decrease in reactant feed to the reactor.
However, the improvement in performance is not as dramatic because the
feed flow has a direct effect on the reactor temperature and cannot be cor-
rected in time by the jacket temperature controller. The improvement in con-
trol is due to the faster response of the reactor temperature to controller
output in the cascade scheme. Another reason that the performance improve-
ment is not as dramatic for the feed flow disturbance is the inverse response of
the temperature to the feed flow. This is because the reactants are colder than
the reactor and the decrease in the flow of reactants causes an immediate rise
in temperature, but the decrease in flow of reactants also causes a decrease in
reactant concentration that eventually results in a decrease in reaction rate
and consequently in temperature.
Figure 7-6. Cascade Control of Reactor Inlet Composition and Synthesis Loop
Pressure in the Ammonia Process
SP SP
RC
Vent 2 AC AT
1 1
Air SP
Compressor FC
2
SP
FT AC
2
Air Synthesis
Gas
AT
2 Compressor
SP Ammonia
FC Synthesis
1
Reactor
FT
Reforming
Natural Process
SP
Gas PC
PT 4
SP
RC
3
SP Ammonia
CO2
SP FC Product
FC
4
3
FT FT
Steam Purge
The objective is to control the hydrogen to nitrogen ratio (H/N) of the mixture
entering the synthesis reactor at its optimum value (about 2.85 for a slight
excess of nitrogen). The primary controller (AC 1) receives the measurement
of the composition at the reactor inlet from a very accurate analyzer (AT 1).
The output of the primary controller adjusts the set point on the secondary
controller (AC 2). The secondary controller receives the measurement of the
composition of the fresh feed from a fast and inexpensive analyzer (AT 2),
usually a simple thermal conductivity detector, and its output adjusts the
ratio of air to natural gas through ratio controller (RC 2). The ratio controller,
in turn, adjusts the set point of the process air flow controller (FC 2). To pre-
136 Tuning of Industrial Control Systems, Third Edition
vent from throttling the air compressor, flow controller (FC 2) adjusts a valve
on a vent on the compressor discharge.
This example illustrates the point made earlier about the secondary measure-
ment not having to be accurate but having to be fast and consistent. Inaccu-
racy in the secondary measurement is corrected by the integral mode of the
primary controller. On the other hand, the measurement of the primary con-
troller can be slow, but it must be accurate. Disturbances in the reforming pro-
cess are handled quickly by the secondary controller, before they have a
chance to affect the primary controlled variable.
Figure 7-6 also shows a pressure-to-flow cascade loop for the control of the
pressure in the synthesis loop. In this cascade the primary controller is the
pressure controller (PC) and the secondary controller is the purge flow con-
troller (FC 4). The purge is a small stream removed from the loop to avoid the
accumulation of inert gases (e.g., argon and methane) and the excess nitrogen.
Although both cascade control loops of Figure 7-6 could be carried out with
analog controllers, computer control offers an unexpected virtue to this
scheme: patience. For example, in one installation where the pressure control
scheme was carried out with analog controllers, the primary controller was
operated on Manual because it was swinging the purge flow all over its range.
This was because the time constant of this loop is about one hour. A digital
controller with a sampling interval of 5 minutes and an integral time of 45
minutes was able to maintain the pressure at its optimum set point on the
same installation.
To illustrate the problem of cascade reset windup, consider the start-up of the
jacketed reactor of Figure 7-1. Both controllers are initially in Manual, with the
Tuning Cascade Control Systems 137
cooling water valve closed and the steam valve manually opened to bring the
reactor up to the operating temperature, 104C (see Figure 7-7). The jacket
temperature transmitter, TT 2, has a range of 40 to 115C, and the steam con-
denses at 110C, which is the value of the jacket temperature when the steam
valve is closed and the cascade control system is initialized and switched to
Automatic. This is done before the reactor temperature reaches its 104C set
point, say when it reaches 100C.
Following the bumpless transfer procedure of the control system, the output
of the primary controller is initialized to the measured temperature of the sec-
ondary controller, 110C. At this time the jacket temperature begins to drop
because the steam has been turned off and the reactor is at the lower tempera-
ture of 100C, while the reactor temperature is rising because of the heat of the
reaction. For the time that the reactor temperature is between 100 and 104C
(its set point), the control situation is as follows:
The secondary controller sees a jacket temperature below its set point
(110C) and calls for the cooling water valve to remain closed.
The primary controller also sees its temperature below set point and
calls for an increase in the jacket temperature set point above the cur-
rent 110C value.
Most computer and DCS controllers detect that the secondary controller out-
put is limited or clamped at the closed position and prevent the primary
controller from increasing its output since this would only call for the closing
of the coolant valve, which is already closed. Does this logic prevent the cas-
cade control system from winding up? Let us see what happens next.
Notice that a gap has been created between the set point of the secondary con-
troller, clamped at 110C, and its measured temperature, the jacket tempera-
ture that drops to the reactor temperature as soon as the steam is turned off.
As the reactor temperature crosses its set point of 104C, the primary control-
ler starts decreasing the set point of the secondary controller to bring the tem-
perature down, but the coolant valve will not open until the set point of the
secondary controller drops below its measured temperature; that is, until the
gap mentioned earlier is overcome. Since the set point of the secondary con-
troller will change at a rate controlled by the integral time of the primary con-
troller, it takes a long time for the coolant valve to start to open and the reactor
temperature overshoots its set point badly, the common symptom of reset
windup. The situation continues as the coolant valve is driven from closed to
open and back again, as the oscillations of the dashed lines in Figure 7-7 show.
As you can see, the saturation or clamp limit detection system could not
avoid reset windup in this case.
Figure 7-7. Oscillatory Behavior Caused by Reset Windup in the Cascade Control
of an Exothermic Reactor Startup (Dashed Lines) and Solution Using Reset
Feedback on the Master Controller (Continuous Lines)
Reset Feedback
An elegant and effective way to protect against cascade reset windup is the
use of the reset feedback feature on the primary controller. In the cascade
Tuning Cascade Control Systems 139
Mk = FBk + Mk
where:
FBk = the reset feedback variable, in this case the measured variable of
the secondary loop
By using this formula to update the set point of the secondary controller at
every processing of the primary controller, there is no possibility of windup
because the primary controller will call for an increase or decrease of the sec-
ondary variable from its current measured value, not from its set point. This
eliminates the gap between the secondary process variable and its set point
when the controller output is saturated.
The result of the use of the reset feedback feature on the cascade temperature
control of the jacketed exothermic reactor is shown by the continuous lines in
Figure 7-7. Although the initial overshoot of the temperature cannot be elimi-
nated because the coolant valve starts in the closed position, further oscilla-
tions are eliminated.
The use of the reset feedback approach requires that the secondary loop be
sampled more frequently than the primary loop, and that the secondary con-
troller have integral mode. Otherwise, any offset in the secondary controller
will cause an offset in the primary controller, even if the primary controller
has integral mode.
140 Tuning of Industrial Control Systems, Third Edition
7-5. Summary
This chapter presented the cascade control scheme; that is, the cascading of a
primary controller to a secondary controller to improve control performance.
The discussion included the reasons for using cascade control, the selection of
modes for the secondary controller, and the procedure for tuning cascade con-
trol systems. It also looked at cascade reset windup and ways to protect
against it.
Review Questions
7-2. What is the main requirement for a cascade control system to result in
improved control performance? What is required of the sensor for the
secondary loop?
7-3. Are the tuning and selection of modes different for the primary control-
ler in a cascade control system than for the controller in a simple feed-
back control loop? Explain.
7-4. What is different about the secondary controller in a cascade control sys-
tem? When should it not have integral mode? If the secondary is to have
derivative mode, should it operate on the process variable or on the error
(deviation)?
7-5. In what order must the controllers in a cascade control system be tuned?
Why?
7-6. What are the two major difficulties with using temperature as the pro-
cess variable of the secondary controller in a cascade control system?
How can they be handled?
7-7. Why is pressure a good variable to use as the secondary variable in cas-
cade control? What are the two major difficulties with using pressure as
the secondary variable?
Tuning Cascade Control Systems 141
7-8. What is the relationship between the processing frequencies of the pri-
mary and secondary controllers in a computer cascade control system?
7-9. How can reset windup occur in a cascade control system? How can it be
avoided?
8
Feedforward and Ratio
Control
This chapter presents the design and tuning methods of feedforward and ratio
control strategies. Along with cascade control, these strategies can be classi-
fied as multiple input, single output (MISO) because they require more than
one process measurement but only one final control element (usually a control
valve) because there is only one control objective.
143
144 Tuning of Industrial Control Systems, Third Edition
Chapter 4 showed that some feedback loops are more controllable than others
and that the uncontrollability measure of a feedback loop is the ratio of the
dead time to the time constant. When this ratio is high, on the order of unity
or greater, feedback control cannot prevent disturbances from causing large
deviations of the process variable from its set point. It is then that the strate-
gies of feedforward and ratio controlratio control being the simplest form of
feedforward controlcan improve control performance the most.
The feedback controller can be tuned on-line, by trial and error, so that
a model of the process is not needed to implement it.
The integral mode of the controller computes the value of the controller
output OP required to keep the process variable PV at its set point SP.
Opposite these very desirable characteristics there are two undesirable ones:
Disturbances cause the process variable to deviate from its set point
before the controller can take action.
These problems are significant in process systems because of the long time
delays involved, sometimes of the order of hours. One remedy to these prob-
lems is feedforward control.
G2
SP OP - PV
Feedback +
Controller G1
+
-
Sensor
146 Tuning of Industrial Control Systems, Third Edition
Figure 8-2 shows a block diagram for feedforward control. The technique con-
sists of measuring the disturbance, U, instead of the process variable being
controlled, PV. Corrective action begins as soon as the disturbance enters the
system and in theory, can prevent any deviation of the process variable from
its set point. However, this requires an exact model of the process and its
dynamics, plus exact compensation for all possible disturbances. The set
point element 1/G1 of Figure 8-2 provides for calibrated adjustment of the
set point and seldom includes any dynamic compensation.
The feedforward element G2/G1 of Figure 8-2 simulates the effect of the dis-
turbance on the process variable (block G2) and compensates for the lags and
delays on the output variable (block G1). Notice that the signals always travel
forward; that is, there is no loop in the diagram, so the feedforward controller
cannot introduce or prevent instability in the process response.
U
G2/G1
G2
SP - +
OP + PV
1/G1 + G1
Feedforward-Feedback Control
It is seldom practical to measure all the disturbances that affect the process
variable. A more reasonable approach is to measure only those disturbances
that are expected to cause the greatest deviations in the process variable and
handle the so-called minor disturbances by adding feedback trim to the
feedforward controller. Figure 8-3 shows a block diagram for a feedforward-
feedback control system. Note that the feedback controller takes the place of
Feedforward and Ratio Control 147
the set point element of Figure 8-2, and only the feedforward element is neces-
sary in the combined control scheme. A feedforward element is required for
each disturbance measured.
U
G2/G1
G2
- +
SP Feedback OP + PV
Controller + G1
+
-
Sensor
When the outputs of the feedforward and feedback controllers are summed,
as in Figure 8-3, the presence of the feedforward controller does not affect the
response of the loop to inputs other than the measured disturbance, thus the
feedback controller tuning does not have to be adjusted because of the instal-
lation of the feedforward controller.
Economics dictates that only those disturbances that are frequent enough and
important enoughin regards to their effect on product quality or safety, or
for similar considerationsshould be measured and compensated for with a
feedforward controller.
The feedback controller takes care of those disturbances that are not
important enough to be measured and compensated for.
148 Tuning of Industrial Control Systems, Third Edition
The feedforward controller does not have to compensate exactly for the
measured disturbances since any minor errors in the model are
trimmed off by the feedback loop, hence the term feedback trim.
Ratio Control
The diagram of the control of the ammonia process in Figure 7-6 shows two
ratio controllers to compensate for variations in the natural gas flow to the
process.
Some control engineers prefer to calculate the ratio by dividing the manipu-
lated flow by the wild flow and then controlling the ratio with a feedback con-
troller, as in RC in Figure 8-5. This alternative has the disadvantage of creating
a very nonlinear feedback control loop; note that the gain of the feedback loop
in Figure 8-5 is inversely proportional to the wild flow, which is the major dis-
turbance. The ratio controllers in some computer and distributed control sys-
tems display the calculated ratio, but do not use it for control. Instead, the
output is calculated by multiplying the input or wild flow by the ratio set
point, as in Figure 8-4.
Feedforward and Ratio Control 149
SP
RC
Steam SP
SP
FT FC
TC
OP
Fs
PV
FT TT
F
Process T
fluid
Steam
trap
Condensate
Based on the block diagram of Figure 8-2, the feedforward controller and the
process constitute two parallel paths between the disturbance U and the pro-
cess variable PV. A simple linear feedforward model assumes that the process
variable response is the sum of its separate responses to the output variable
OP and to the disturbance U:
PV = G1(OP) + G2(U)
Wild flow
FT
(B/A)set
A
SP
B/A
FY RC
%
B
Manipulated FT
flow
1 G
OP = ------- SP ------2- U (8-1)
G1 G1
This is the design equation for the feedforward controller having set point SP
and disturbance U as inputs and output variable OP as output. Equation 8-1
provides the design formulas for both the set point and feedforward elements
of Figure 8-2:
Feedforward element:
G
G F = ------2-
G1
Feedforward and Ratio Control 151
When feedback trim is used, as in Figure 8-3, only the feedforward element is
needed, since the feedback controller takes the place of the set point element.
When the process elements G1 and G2 are modeled with simple single-lag-
plus-dead-time (SLPDT) models, the feedforward controller can be built out
of standard algorithms available in most commercial process control pro-
grams. The feedforward controller then consists of three elements:
with:
K
Gain = ------2-
K1
Lead of
Lead Lag = --------------------------1-
Lag of 2
where:
Although the feedforward controller of Equation 8-2 results from simple sin-
gle-lag process models, there is no incentive to use more complex dynamic
compensation terms. For example, use of process models with more than one
lag would call for a compensator with additional parameters than the lead-lag
unit, making it harder to tune while offering little improvement in perfor-
mance over a well-tuned lead-lag unit.
152 Tuning of Industrial Control Systems, Third Edition
The dead time compensator of Equation 8-3 can only be realized when the
dead time between the disturbance and the process variable is longer than the
dead time between the manipulated variable and the process variable. Other-
wise, it would call for the feedforward correction to start before the distur-
bance takes place, which is obviously not possible.
The dead time compensator requires the memory of digital devices (comput-
ers and microprocessors) for its implementation. The dead time compensator
can often be left out because the lead-lag unit can be tuned to provide all of
the required dynamic compensation, thus simplifying the tuning task. In gen-
eral, the dead time compensator should only be used when the lead-lag unit
cannot do the job by itself.
Of the three terms of the feedforward controller of Equation 8-2, the gain is
always required and the dynamic compensators are optional. When only the
gain is used, the feedforward controller is called a static compensator.
Gain Adjustment
The adjustment of the feedforward gain can be carried out with the feedback
controller in Manual or Automatic. If it is done with the feedback controller in
Manual, when the gain is not correct, the process variable will deviate from its
set point after a sustained disturbance input. The gain can then be adjusted
until the process variable is at the set point again. Because of process nonlin-
earities, the required feedforward gain may change with operating conditions,
thus exact compensation may not be possible with a simple linear controller.
The one thing to remember when tuning the feedforward gain is that it is nec-
essary to wait until the system reaches steady state before making the next
adjustment.
Figure 8-6 shows the response of the lead-lag unit to a step change in its input
for both the lead being longer than the lag and for the lag being longer than
the lead, assuming in each case that the gain is unity. The initial change in the
output of the lead-lag unit is always equal to the ratio of the lead to the lag, so
that there is an initial overcorrection when the lead is longer than the lag and a
partial correction when the lag is longer than the lead. In either case, the out-
put approaches the steady state correction asymptotically, at a rate deter-
mined by the lag time constant.
Figure 8-7 shows the response of the lead-lag unit to a ramp input, both for
the lead longer than the lag and the lag longer than the lead, assuming unity
gain. The figure shows where the terms lead and lag come from: the out-
put of the lead-lag unit, after a transient period, either leads the input ramp by
the difference between the lead and the lag or lags it by the difference between
the lag and the lead. The ramp response is more typical than the step response
to the type of inputs provided by the disturbances in a real process. The ramp
response is similar to the response to the rising and dropping portions of
slow-oscillating disturbances.
With the responses to step and ramp inputs in mind, tuning the lead-lag unit
becomes a simple procedure, as follows:
1. Decide by how much to lead or lag the feedforward correction to the dis-
turbance; this fixes the difference between the lead and the lag.
154 Tuning of Industrial Control Systems, Third Edition
3.5
3
Output
2.5
Lead = 2xLag
2
Input
1.5 Output
Lead = 0.5xLag
1
0.5
0
0 5 10 15 20 25 30 35 40 45
2. Select the ratio of the lead to the lag based on how much to amplify or
attenuate sudden changes in the disturbance inputs. For example, suppose
it is desired to lead the disturbance by one minute; a lead of 1.1 minutes
and a lag of 0.1 minutes give an amplification factor of 1.1/0.1 = 11, while
a lead of 3 minutes and a lag of 2 minutes give an amplification factor of
only 3/2 = 1.5. If the disturbance is noisy; for example, in the case of a
flow, the second choice is preferred since it results in less amplification of
the noise.
Although it is possible to have a lag with zero lead, it is not possible to have a
lead without a lag. The ratio of the lead to the lag should not be greater than
10. When a net lag is required, the lead can usually be set to zero, simplifying
the tuning task.
Feedforward and Ratio Control 155
120
100
Net lead
80
Output
Lead > Lag Net lag
60
40 Input
20 Output
Lead < Lag
0
0 5 10 15 20 25 30 35 40 45
LD
Y k = Y k 1 + ( 1 a ) ( X k 1 Y k 1 ) + --------- ( X k X k 1 ) (8-4)
LG
where:
Equation 8-4 is for a unity gain. If the gain is different from unity, it can be
applied to the signal before or after the lead-lag calculation.
Dead time compensation is not possible when the dead time is negative
because it would require taking action ahead of the disturbance, so
when the dead time is negative it should be added to the lead term in
the lead-lag unit and only the corrected lead-lag unit should be used.
Instead of using a lead in the lead-lag unit and a dead time compensa-
tion term, it is simpler to just subtract the dead time from the lead and
use only the corrected lead-lag unit.
When the dead time in the dead time compensation term is longer than
the lead in the lead-lag unit, it is simpler to subtract the lead from the
dead time and use a lag without lead and a corrected dead time com-
pensation term.
In other words, dead time compensation should be used only when a lag
without a lead would cause the feedforward correction to take place too soon.
Yk = Xk-N
Feedforward and Ratio Control 157
where N is the number of samples of dead time and unity gain is assumed.
The dead time compensator is easy to tune, since it only has one dynamic
parameter, the number of samples of dead time N.
4
Input
3
Dead Time = 20 min
0
0 5 10 15 20 25 30 35 40 45
158 Tuning of Industrial Control Systems, Third Edition
1. State the control objective; that is, define which process variable needs to
be controlled and what its set point is. It is useful to write the objective in
the form:
The set point should be adjustable by the operator and should not be a
constant.
3. Select the output variable, that is, the variable to be adjusted by the feed-
forward controller. When the feedforward controller is cascaded to a sec-
ondary controller, the output variable should be defined as the set point of
the secondary controller; for example, the flow of the manipulated stream
instead of the valve position.
4. From basic principles, usually material and energy balances, write the for-
mulas relating all the variables defined in the first three steps. Keep them
Feedforward and Ratio Control 159
as simple as possible. Solve for the output variable so that it can be calcu-
lated from the measured disturbances and the set point of the process
variable. The resulting formula or formulas constitute the design equa-
tion(s) to be programmed into the computer for on-line execution. Cau-
tion: the formula must use the set point of the process variable and not its
measured value.
6. Introduce the feedback trim, if any, into the design equation. This is done
by grouping unknown terms and unmeasured disturbances as much as
possible and letting the output of the feedback controller adjust the group
of terms that is expected to vary the most. A simple and effective approach
is to have the output of the feedback controller adjust the set point of the
feedforward controller.
1. Control objective:
To = Toset (8-5)
2. Measured disturbances:
W, the flow through the exchanger, kg/h
Ti, the inlet temperature, C
3. Manipulated variable:
F, steam flow controller set point, kg/h
where:
SP
Feedforward Fset
Controller Toset
Steam SP
SP
FT FC
TC
OP
F
Ti W PV
TT FT TT
Process To
fluid
Steam
trap
Condensate
6. The need for feedback trim is determined by considering how much the
unknown terms in the design formula are expected to vary. The three
unknown terms are the physical properties, Cp and Hv, and the heat loss
rate, QL. The three can be lumped together by assuming that the heat
loss rate is proportional to the heat transfer rate:
QL = (1 - )FHv (8-7)
162 Tuning of Industrial Control Systems, Third Edition
set Cp set
F = ----------- ( T o T i )W (8-8)
Hv
Notice that the outlet temperature in the formula has been replaced by
its set point; that is, the control objective, Equation 8-5, has been
substituted into the design formula to ensure that it is enforced by the
feedforward controller. In modern computer control systems it is
possible to retrieve the set point from the feedback controller to use in
the feedforward calculation, so that only one set point has to be entered
by the operator. This is an important design requirement.
All the unknowns of the model have been lumped into a single
coefficient, Cp/Hv, and it would seem natural for the feedback trim
controller to adjust this coefficient to correct for variations in the
specific heat Cp, the steam latent heat of condensation Hv, and the
heater efficiency . However, these parameters are not expected to vary
much, thus it would be undesirable for the feedback trim controller to
control by adjusting a term that is not expected to vary. A better control
system structure results if the feedback controller output is made to
adjust the set point of the feedforward controller or equivalently, the
product of the unknown coefficient and the set point, as follows:
set Cp
F = OP -----------T W
H v i
where:
Cp F
----------- = ------------------------------
Hv W ( To Ti )
8. Figure 8-10 shows the instrumentation diagram for the feedforward con-
troller. In some computer control systems, the multiplier may be carried
out as a ratio controller, with the ratio being set by the adder, which
combines the feedback controller output and the inlet temperature cor-
rection.
Tune the lead-lag units for the steam heater feedforward controller of the
preceding example.
TT TT Toset FT
Ti To SP W
Lead-Lag Lead-Lag
TC 2
1
+
-
Adder
Toset - Ti
Multiplier
Fset
lead-lag unit reduces the deviation of the outlet temperature to about one-
third the deviation of the simple feedback controller and about two-thirds
that of the static feedforward controller.
(a)
(b)
(c)
8-5. Summary
References
Review Questions
8-1. Why isnt it possible to have perfect control; that is, the process variable
always being equal to the set point, with feedback control alone? Is per-
fect control possible with feedforward control?
8-2. What are the main requirements of feedforward control? What are the
advantages of feedforward control with feedback trim over pure feedfor-
ward control?
8-3. What is ratio control? What is the control objective of the air-to-natural
gas ratio controller in the control system sketched in Figure 7-6 for the
ammonia process? Which are the measured disturbance and the manipu-
lated variable for that ratio controller?
Flue gas
W Ti
FT TT
Process To
Stream TT
Fset
SP
F
FC Fs
FT
FT
Auxiliary
Main Fuel Air Fuel
9
Multivariable Control
Systems
169
170 Tuning of Industrial Control Systems, Third Edition
When two or more feedback loops are installed on a process or unit operation
(e.g., distillation column, evaporator, etc.), the possibility arises of interaction
between the loops. This means that each process variable is affected by more
than one controller output as shown in Figure 9-1, where in controlling the
total flow and concentration out of a catalyst blender, both process variables
are affected by each of the two controller outputs, which are the flows of the
concentrated and dilute inlet streams. The problem that arises is known as
loop interaction. Since multiple control objectives are involved, the problem can
also be viewed as the design of a multivariable control system.
SP
AC SP
OP1 FC
x
F1x1 F
OP2 AT FT
F2x2
SP2 PV2, to produce the two controller outputs. Signals SP1 and SP2 repre-
sent the set points of the loops. In the diagram of Figure 9-2 each of the four
process blocks, G11, G12, G21 and G22, includes the gains and dynamics of the
final control elements (valves), the process and the sensor/transmitters. For
simplicity the disturbances are not shown.
SP1 - +
Controller OP1 PV1
G11
+ 1
+
G12
G21
SP2 +
Controller OP2 + PV2
G22
+ 2
-
To look at the effect of interaction, assume that the gains of all four process
blocks are positive; that is, an increase in each controller output results in an
increase in each of the process variables. Let us follow the sequence of events
plotted in Figure 9-3:
3. The decrease in OP2 also causes, through interaction block G12, a decrease
in process variable PV1, so that the net change in PV1 is smaller than the
initial change. Note that this initial change is the only change that would
take place if there were no interaction, or if controller 2 were kept in
Manual.
The difference between the initial change and the net change in PV1 is the
effect of interaction. It depends on the effect that OP1 has on PV2 (G21), the
effect that OP2 has on PV2 (G22)which determines the necessary corrective
action on OP2and the effect that OP2 has on PV1 (G12). Note also that the
steady-state effect of interaction depends only on the process gains, not on the
controller tuning, provided that controller 2 has integral mode.
Figure 9-3. Effect of Interaction for a 2x2 Control System with all Four Gains
Positive
b
OP1
b
PV1 SP1
a a
a b b
OP2 PV2
SP2
a
Time
The authors invite you to verify that a step in OP2, followed by closing control
loop 1, has the same effect on PV2at least qualitativelyas the effect just
observed on PV1. It will be shown shortly that the relative effect of interaction
is quantitatively the same for control loop 2 as it is for control loop 1.
Multivariable Control Systems 173
In the case just analyzed, all four process gains were assumed positive (direct
action). The effect of interaction was in the opposite direction as the direct (ini-
tial) effect of the step change, resulting in a net change that was smaller than
the initial change. This type of situation, in which the two loops fight each
other, is known as negative interaction. Note that it is possible for the
effect of interaction to be greater than the initial effect, in which case the net
change will be in the opposite direction as the initial change. Here we could
say that the wrong loop wins the fight, a situation that results from incorrect
pairing of the loops, as shall be shown shortly. You can easily verify that if any
two of the process gains were positive, and the other two were negative, the
interaction would also be negative.
If one of the four process gains has a sign opposite to that of the other three,
the effect of interaction would be in the same direction as the direct action and
the net change would be larger than the initial change, as you can also verify.
This is the case of positive interaction, in which the two loops help each
other.
The following sections look at two ways to approach the problem of loop
interaction:
The first step in the design of a control system for a process is usually the
selection of the control loops; that is, the selection of those variables that must
be controlled and of those variables that are to be manipulated to control
them. This pairing task has traditionally been performed by the process engi-
neer and based mostly on intuition and knowledge of the process. Fortu-
nately, for a good number of loops, intuition is all that is necessary. However,
when the interactions involved in a system are not clearly understood and the
intuitive approach produces the wrong pairing, control performance is
poor. The expedient solution is then to switch the troublesome controllers to
Manual which, as pointed out in the preceding section, eliminates the effect of
interaction. The many controllers operating in Manual in control rooms
throughout industry are a testimony of failures to correctly pair the variables
in the system. Each one is a failure of an attempt to apply automatic control.
Open-loop Gains
changes in process variables PV1 and PV2 are measured, the open-loop gains
can be calculated:
Change in PV
K 11 = ---------------------------------------1-
Change in OP 1
(9-1)
Change in PV
K 21 = ---------------------------------------2-
Change in OP 1
Similarly, when a change is applied to OP2, keeping OP1 constant, the other
two open-loop gains can be calculated:
Change in PV
K 12 = ---------------------------------------1-
Change in OP 2
(9-2)
Change in PV
K 22 = ---------------------------------------2-
Change in OP 2
The open-loop gains can be determined from the steady-state equations or the
computer simulation programs used to design the plant.
There is a natural tendency to try to use the open-loop gains in the pairing of
the variables. However, it is immediately apparent that PV1 and PV2, and OP1
and OP2 do not necessarily have the same dimensions. Thus, attempting to
compare open-loop gains would be like trying to decide between buying a
new sofa or a new house. To overcome this problem, Bristol1 proposes to com-
pute relative gains that are independent of dimensions.
Closed-loop Gains
Because of interaction, the effect of OP1 on PV1 is different when the other
loop is closed than when it is opened, as discussed in the previous section.
This requires the definition of the closed-loop gains K11', K21', K12' and K22'.
These are defined exactly in Equations 9-1 and 9-2, but with the changes in
PV1 determined with PV2 kept constant, and the changes in PV2 determined
with PV1 kept constant. For example, to determine K11', a change is made in
OP1 and the change in PV1 is measured while a feedback controller with inte-
gral mode controls PV2 by adjusting OP2.
176 Tuning of Industrial Control Systems, Third Edition
However, closed-loop tests are not needed because the closed-loop gains can
be computed from the open-loop gains previously defined. For example,
when both OP1 and OP2 change, the total change in PV1 can be estimated by
the sum of the two changes:
and similarly for the total change in PV2. Now, if PV2 is kept constant, its
change is zero:
Solving for the change in OP2 required for PV2 to remain constant:
K 21
- ( Change in OP 1 )
Change in OP 2 = --------
K 22
K 12 K 21
- ( Change in OP 1 )
Change in PV 1 = K 11 ------------------
K 22
The bracketed expression is then the closed-loop gain K11'. The closed-loop
gains for each of the other three pairings can be similarly derived.
K
ij = -------ij-
K ij'
where ij is the relative gain for the pairing of process variable PVi with con-
troller output OPj.
Multivariable Control Systems 177
The formulas of Equations 9-3 can be used to compute the relative gains for
any 2x2 system:
K 11 K 22
11 = 22 = -------------------------------------------
-
K 11 K 22 K 12 K 21
(9-3)
K 12 K 21
12 = 21 = -------------------------------------------
-
K 12 K 21 K 11 K 22
It makes sense that the interaction measure for the PV1-OP1 pair is the same as
for the PV2-OP2 pair since they represent one option in the 2x2 system, the
other option being PV1-OP2 and PV2-OP1.
The relative gains are dimensionless and can therefore be compared to one
another. To minimize the effect of interaction, the process variables and con-
troller outputs are paired so that the relative gain for the pair is closest to
unity. This results in the least change in gain when the other loop of the pair is
closed. Note that for the case of no interaction, the open-loop gain is equal to
the closed-loop gain, and the relative gains are 1.0 for one pairing and 0.0 for
the other.
In the blender of Figure 9-1, a change of 5 kg/h in F1, the dilute inlet stream,
results in a steady-state increase of 5 kg/h in F, the outlet flow, and a
decrease of 0.5% in x, the outlet mass% of solute. A change of 2 kg/h in F2,
the concentrated inlet stream, results in a steady-state increase of 2 kg/h in F
and an increase of 0.8% in x. Determine the relative gains and pair the flow
controller, FC, and mass fraction controller, AC, so as to minimize
interaction.
This means that for the pair F1 with F and F2 with x, the steady-state gain of
each loop increases to 1/0.8 = 1.25 (a 25% change) when the other loop is
closed, while for the pair F1 with x and F2 with F, the gain of each loop
increases by a factor of 1/0.2 = 5 (a 400% change) when the other loop is
closed! Obviously the first pairing is significantly less sensitive to interac-
tion than the second.
Equation 9-3 can be used to compute the relative gains for any 2x2 control sys-
tem. For systems with more than two process and controller outputs, the open
loop gain of each loop is determined with all the other loops opened, and the
closed loop gain implies that all the other loops are closed. The relative gain
for each process variable, controller output pair is still defined as the ratio of
the open-loop gain to the closed-loop gain for that pair.
The calculation of the relative gains involves the inversion of the matrix of
open-loop gains, which is K in Equation 9-4. It is therefore helpful to use a
computer and canned programs to perform the following matrix operations:
1. The relative gains are not only non-dimensional, but are also normalized
in the sense that the sum of the gains of any row or column of the matrix is
unity. You can verify this fact for the 2x2 by adding the relative gain for-
mulas for each pairing, that is, 11 + 12 = 1. This property also applies to
systems with more than two process and controller outputs.
2. For the 2x2 system, when the two loops help each other (positive interac-
tion), the relative gains are between 0 and 1; conversely, when the two
loops fight each other (negative interaction), one set of relative gains is
greater than unity and the other set is negative. Notice that a negative rel-
ative gain means that the net action of the loop reverses when the other
loop is opened or closed, a very undesirable situation.
3. For a system with more than two control objectives, the concept of positive
and negative interaction must be applied on a pair by pair basis. If the rel-
ative gain for a process variable and controller output pair is positive and
less than unity, the interaction is positive; that is, that pair is helped by
the interaction of all the other loops. On the other hand, if the relative gain
for a pair is greater than unity or negative, the interaction is negative. In
other words, the combined action of all other loops causes a change in the
process variable that is in the opposite direction as the direct change
caused by the controller output in the pair.
Consider the blender of Figure 9-1, where the objectives are to control the
composition x and flow F of the product stream by adjusting the positions
of the control valves on the two feed streams. Which of the two controllers
should be paired to which valve to minimize the effect of interaction? The
relative gains can be used to determine this. (Note: Although ratio control
should be used here, this still leaves the question of which flow should be
ratioed to which, and the answer to our original question will also answer
this one. In fact, the ratio controller is really a form of decoupling here.)
180 Tuning of Industrial Control Systems, Third Edition
Unlike the specific numerical solution that was developed in Example 9-1, a
general solution for the blender will be developed here. To do this, the con-
servation of mass and the conservation of solute are used to develop formu-
las for the open-loop gains.
Conservation of mass: F = F1 + F2
F1 x1 + F2 x2
Conservation of solute: x = -----------------------------
-
F1 + F2
F2 ( x1 x2 ) F1 ( x2 x1 )
K x1 = ---------------------------
-K K x2 = ---------------------------
-K
2 v1 2 v2
( F1 + F2 ) ( F1 + F2 )
where Kv1 and Kv2 are the valve gains in (kg/h)/fraction valve position.
Next substitute the open-loop gains into the formulas for the relative gains,
Equations 9-3. A little algebraic manipulation produces the following gen-
eral expressions for the relative gains:
F1 F2
F1 = x2 = -----------------
- F2 = x1 = -----------------
-
F1 + F2 F1 + F2
In words, the pairing that minimizes interaction has the flow controller
adjusting the larger of the two flows and the composition controller adjust-
ing the smaller of the two flows. If a ratio controller were to be used, the
smaller flow would be ratioed to the larger flow, with the flow controller
adjusting the larger flow and the composition controller adjusting the ratio.
It could easily be shown that the ratio controller decouples the two loops so
that a change in the product stream flow does not affect the composition.
Note that the valve gains Kv1 and Kv2 do not affect the relative gains. This is
why they were not considered in Example 9-1.
Multivariable Control Systems 181
For most processes, the relative gains tell all that needs to be known about
interaction. They are determined from the open-loop steady-state gains,
which are easy to determine by either on-line or off-line methods. However,
in systems with negative interaction, the pairing recommended by relative
gain analysis may not result in the best control performance because it does
not consider the dynamic response. This is illustrated in the following
example.
Figure 9-4 shows a sketch of a distillation column with five controller out-
puts and five process variables. The column separates a 50% mixture of ben-
zene and toluene into a distillate product with 95% benzene and a bottoms
product with 5% benzene. It is desired to maintain the compositions of the
distillate and bottoms products at their set points. In a distillation column,
temperature can provide an indirect measurement of composition, so the
two temperature controllers (TC 1 and TC 2) control the composition of the
two products by inference. Three secondary objectives are to maintain the
vapor balance by controlling the column pressure (PC), and the liquid bal-
ances by controlling the levels in the accumulator drum (LC 1) and column
bottom (LC 2). The five controller outputs adjust the flow rates of the two
products, the reflux flow, the steam flow to the reboiler, and the cooling rate
of the condenser.
SP Condenser
PC PT SP
LC
LT
1
TT
TC
Feed Reflux Distillate
1
SP
SP
TT TC
2
SP
LC LT
2 Steam
Reboiler Bottoms
SP Condenser
PC PT SP
LC
LT
1
TT
TC
Feed Reflux l
Distillate
1
SP
SP
TT TC
2
SP
LC LT
2 Steam
Reboiler Bottoms
Reflux Steam
TC 1 -2.85 1.16
TC 2 -0.438 2.53
Reflux Steam
TC 1 1.08 -0.08
TC 2 -0.08 1.08
The recommended pairing is the same as the intuitive one: control the over-
head composition with the reflux flow and the bottoms composition with the
184 Tuning of Industrial Control Systems, Third Edition
bottoms flow. The interaction is negative, with the gain of each loop decreas-
ing by 8% when the other loop is closed.
SP Condenser
PC PT SP
LC
LT
1
TT
TC
Feed Reflux l
Distillate
1
SP
SP
TC
TT
2
SP
LC LT
2 Steam
Reboiler Bottoms
The sensitivity study on the simulated column gives the following open-loop
gains:
Reflux Bottoms
TC 1 -0.35 -1.05
TC 2 0.07 -1.93
Multivariable Control Systems 185
( 0.35 ) ( 1.93 )
11 = 22 = ----------------------------------------------------------------------------- = 0.90
( 0.35 ) ( 1.93 ) ( 0.07 ) ( 1.05 )
Reflux Bottoms
TC 1 0.90 0.1
TC 2 0.10 0.90
The pairing for this scheme is also the obvious one: top temperature with
reflux and bottom temperature with bottoms product flow, and the relative
gains show about 10% positive interaction; that is, the two loops help each
other, indicated by the relative gains being positive and less than unity.
It would appear, then, from steady-state relative gain analysis that the Direct
Material Balance Control results in positive interaction, which is preferred to
the negative interaction resulting from Energy Balance Control. Unfortu-
nately, the Energy Balance Control scheme was found by simulation to per-
form better in this particular example than the Direct Material Balance Control
scheme. The reason is dynamic interaction, which goes undetected by the rel-
ative gain matrix. For the first scheme the open-loop responses are monotonic;
that is, the temperature stays between its initial value and its final value dur-
ing the entire response. On the other hand, for the second scheme the open-
loop responses exhibit inverse response; that is, the temperature moves in one
direction at the beginning of the response and then moves back to a final
value on the opposite side of its initial value. This causes the feedback control-
ler to initially take action in the wrong direction, degrading the performance
of the control system.
Although relative gain analysis usually results in the pairing of variables that
minimizes the effect of loop interaction, it does not eliminate it. When the rel-
ative gains approach 0.5, the effect of interaction is the same regardless of the
pairing. In the case of negative interaction, when one set of relative gains is
negative and the other is much greater than unity, the proper pairing still pro-
duces significant interaction. The only solution to this problem is to compen-
sate for interaction by designing a decoupler.
SP1 - + + PV1
Controller OP1 G11
+ 1 +
+
D1
G12
D2 G21
SP2 + +
OP2 + PV2
Controller G22
+ 2 +
-
Each of the two decoupler terms, D1 and D2, can be considered to be feedfor-
ward controllers for which the disturbances are the controller output sig-
nals OP1 and OP2. The design of the decouplers is therefore identical to the
design of a feedforward controller presented in Chapter 8.
Multivariable Control Systems 187
G 12
D 1 = --------
- (9-4)
G 11
G 21
D 2 = --------
- (9-5)
G 22
Unlike the feedforward controller, the decoupler forms a part of the feedback
loop and can thus introduce instability into the system. Consider the total
effects that OP1 has on PV1 and that OP2 has on PV2:
Another aspect of decoupling is that, as Equations 9-6 and 9-7 show, two par-
allel paths exist between each controller output and its process variable. For
processes with negative interaction these two parallel paths have opposite
signs, creating either an inverse response or an overshoot in the open-loop
step response of each decoupled loop. It is important to realize, however, that
the parallel paths are not created by the decouplers since they were already
present in the un-decoupled system (in the interaction and direct effects).
It is evident from the design of the decoupler that the steady-state effect of the
decoupler on any one loop is the same as what the integral mode of the other
loops would have if the decoupler were not used. What, then, does the decou-
pler achieve? Basically, through decoupling the effect of interaction is made
independent of whether the other loops are opened or closed. However, prob-
lems may still arise in one loop if the controller output of another loop is
driven to the limits of its range, because the decoupling action is then blocked
by the saturation of the final control element.
Half Decoupling
pling one loop and letting the other loop be affected; this can be achieved by
implementing either D1 or D2 but not both. This is referred to as half decou-
pling. In deciding which decoupler to select, the first consideration may be
which of the process variables is more important to keep at its set point. A sec-
ondary consideration may be the ease with which the dynamic terms of the
decouplers can be implemented.
The two objectives of the control system for the catalyst blender of Figure
9-1 are the control of the product composition and the control of product
flow. Since the blender is full of liquid, the response of the total flow to
changes in each of the input flows is instantaneous, thus the decoupler for
the total flow should not require dynamic compensation. The response of
the product composition should be that of a simple lag with a time constant
equal to the residence time of the tankthe tank volume divided by the
total flow. Since this time constant is the same for the composition response
to either input flow, no dynamic compensation should be required for the
composition decoupler either.
The application of the linear decoupler design formulas, Equations 9-4 and
9-5, results in the following formulas for the signals to the control valves,
assuming that F1 is the largest of the two flows and, for minimum interac-
tion, it is used to control the total flow. This is the pairing determined by
relative gain analysis in Example 9-2:
K v2
- ( OP 2 OP 20 )
M 1 = OP 1 ---------
K v1
F 2 K v1
- ( OP 1 OP 10 )
M 2 = OP 2 ---------------
F K 1 v2
The coefficients correct for the sizes of the two valves and, in the second
formula, for the ratio between the two inlet flows that is required to main-
tain the composition constant. This ratio is a function of the two inlet
190 Tuning of Industrial Control Systems, Third Edition
stream compositions and the product composition set point. If any of these
compositions were to vary, the gain of the decoupler would have to be read-
justed. There is, however, another way to design the decoupler which does
not require readjustment of the parameters when process conditions
change. It consists of using simple process models to set up the structure of
the control system, as discussed next.
The conservation of total mass balance and of mass of solute from Example
9-2 provides the models needed to design the decouplers. Conservation of
total mass indicates that the output of the product flow controller should
manipulate the sum of the two inlet flows, so the output of the flow controller
is assumed to be the total inlet flow, and the smaller flow is subtracted from it
to determine the larger flow:
This formula requires the measurement of the smaller flow and flow control
of the larger flow.
The conservation of solute mass shows that the product composition depends
on the ratio of the flows rather than on any one of the inlet flows. It is then
assumed that the output of the composition controller is the ratio of the
smaller flow to the larger flow, and the smaller flow is calculated as follows:
This formula requires that the smaller flow also be controlled. Figure 9-8
shows the diagram of the resulting control system. In this scheme the ratio
controller keeps the product composition from changing when the total flow
is changed, and the adder keeps the total flow from changing when the com-
position controller takes action. The multivariable control system is therefore
fully decoupled.
The last two design formulas, Equations (9-8) and (9-9), do not show the scale
factors that may be necessary to convert the flow signals into percent of the
Multivariable Control Systems 191
SP
- + OP1 = (F1 + F2)set
Adder FC
F1set OP2 = (F2/F1)set SP
AC
SP
FC SP
F1
RC
FT x
F2set F
F1x1
SP AT FT
FC
F2
FT
F2x2
scales of the flow controllers. The scale factors depend on the spans of the two
flow transmitters rather than on the sizes of the control valves. The flow con-
trollers allow the signals to be linear with flow, and they also take care of
changes in pressure drop across the control valves.
Kc = 3 TI = 6 min
Figure 9-9 shows the responses of the product composition and flow, as
well as the inlet flows, for a step increase of 10 kg/hr in product flow set
point followed by a 2% increase in product composition set point. The con-
tinuous curves plot the responses for simple feedback controllers and the
dashed lines are for the decoupled system. Notice that for the change in
product flow the decoupler immediately changes both flows, keeping the
ratio between the two flows constant and the product composition constant,
while the un-decoupled system must correct the dilute flow to bring the
composition back to set point. Similarly, for the change in product composi-
tion, the decoupler keeps the total flow constant while the un-decoupled
system requires a small delay in correcting the flows, causing the total flow
to dip temporarily.
Figure 9-9. Product Flow and Composition Control of Blender with Decoupler
(Dashed Lines) and with Simple Feedback Controllers (Solid Lines)
It is obvious from the preceding analysis of interacting loops that the interac-
tion is going to affect the response of each loop; that is, the tuning parameters
and manual/automatic state of each loop affects how the other loops respond.
This section shows how to account for the effect of interaction when tuning
each loop in a multivariable control system.
The first thing to do when tuning interacting loops is to prioritize the control
objectives; in other words, to rank the process variables in the order in which
is important to maintain them at their set points. The second thing to do is to
check the relative gain for the most important variable and decide whether it
is necessary to detune the other loops. The principle behind this approach is
that a loosely tuned feedback control loop with low gain and slow integral
behaves as if it were opened; that is, it will make slow enough changes in its
194 Tuning of Industrial Control Systems, Third Edition
The decision on how loosely to tune the less important loops is based on how
different from unity the relative gain is for the most important loop. It is
understood that the controller output for the most important variable has
been selected to make the relative gain for that loop as close to unity as possi-
ble. When there are more than two interacting loops, the tightness of tuning
for each loop will decrease with its rank of diminishing importance.
If all the loops are of equal importance and speed of response, each must be
tuned while the other loops are in Manual. The gain of each loop must then be
adjusted by multiplying the gain obtained when all other loops were opened
by the relative gain for the loop:
where:
Kcij = the controller gain tuned with all the other loops opened
This adjustment accounts for the change in steady-state gain when the other
loops are closed, but it does not account for dynamic effects. If some of the
loops are slower than the others or can be detuned, the relative gains for the
remaining loops must be recalculated as if those were the only interacting
loops; that is, as if the slower or detuned loops were always opened.
The gain adjustment suggested by Equation 9-10 should be sufficient for those
loops with positive interaction since their response remains monotonic when
the other loops are closed. However, the loops with negative interaction may
require retuning after the other loops are closed because the other loops will
cause either inverse or overshoot responses that normally require lower gains
and slower integral than monotonic loops. Note that the formula results in a
gain reduction for the loops with positive interaction and a gain increase for
the loops with negative interaction, assuming that the pairing with the posi-
tive relative gain is always used, as it should be.
When decouplers are used, they must be tuned first and kept active while the
feedback controllers are tuned. Recall that perfect decoupling has the same
effect on a loop as if the other loops were very tightly tuned. For example, in
the blender control system of Figure 9-8, the ratio and mass balance control-
lers must be tuned first and then kept active while the flow and composition
controllers are tuned.
196 Tuning of Industrial Control Systems, Third Edition
Because of the additional dead time of the analysis and the sampling, the
analyzer controller AC tuning changes as follows:
Figure 9-10. Control of Product Composition with the Product Flow Controller in
Automatic (Solid Lines) and in Manual (Dashed Lines)
cusses Model Reference Control and presents an example. For a simple intro-
duction to the mathematics of the DMC scheme see Smith and Corripio3.
corrected. The corrected predicted values from the model are then used to
determine the changes in the manipulated or independent variables that mini-
mize the deviations of the dependent variables from their set points.
Because the different process variables have different units of measure (e.g.,
temperatures, flows, compositions, etc.), their deviations must be weighted in
the function to be minimized. One way this is done is by defining an equal-
concern error for each variable. For example, equal-concern errors in a given
application may be 5C, 200 kg/h, 2 weight%, etc. Weighing the deviations by
the reciprocals of the equal-concern errors normalizes them into deviations of
equivalent magnitude.
9-6. Summary
This chapter dealt with multivariable control systems and their tuning. It
showed the effect that loop interaction has on the response of feedback control
systems and presented two methods to deal with it: Bristols relative gains for
quantitatively determining the amount of interaction and for selecting the
Multivariable Control Systems 199
References
Review Questions
9-1. Under what conditions does loop interaction take place? What are its
effects? What two things can be done about it?
9-2. For any given loop in a multivariable (interacting) system, define the
open-loop gain, the closed-loop gain, and the relative gain (interaction
measure).
9-3. How are the relative gains used to pair process variables and controller
outputs in an interacting control system? What makes it easy to deter-
mine the relative gains? What is the major shortcoming of the relative
gain approach?
9-4. In a 2x2 control system the four relative gains are 0.5. Is there a best way
to pair the variables to minimize the effect of interaction? By how much
does the gain of a loop change when the other loop is closed? Is the inter-
action positive or negative?
9-5. Define positive and negative interaction. What is the range of values of
the relative gain for each type of interaction?
200 Tuning of Industrial Control Systems, Third Edition
9-6. The open-loop gains for the top and bottom compositions of a distillation
column are the following:
Reflux Steam
Distillate Composition 0.05 -0.02
Bottoms Composition -0.02 0.05
Calculate the relative gains and pair the compositions of the distillate
and bottoms to the reflux and steam rates so that the effect of interaction
is minimized.
9-7. The automated showers in the house of the future will adjust the hot and
cold water flows to maintain constant water temperature and flow. In a
typical design the system is to deliver 2.5 gallons per minute (gpm) of
water at 110F by mixing water at 170F with water at 80F. Determine
the open-loop gains, the relative gains and the preferred pairing for the
two control loops. Hint: the solution to this problem is identical to that of
Example 9-2.
9-8. Design a decoupler to maintain the temperature constant when the flow
is changed in the shower control system of the preceding exercise.
Dynamic effects can be neglected.
10
The Auto-tuner
Application
This chapter will describe the functionality and operation of the auto-tuner
application that is built into most process control software packages on the
market at this time. Knowing how to properly set up and use the application
is essential to obtaining acceptable results from the auto-tuner.
201
202 Tuning of Industrial Control Systems, Third Edition
10-1. Operation
The auto-tuner is simply a tool that allows the engineer or technician to per-
form the tuning procedures discussed in this book in a timely, more efficient
manner. However, the application must be set up for the task at hand. This
requires knowing how to configure the application for each loop that needs to
be tuned.
Once the application has been configured properly, it will perform satisfacto-
rily. The auto-tuner will perform the step test, make the necessary measure-
ments and solve the calculations depending on the selected algorithm. There
are many algorithms to choose from, including Modified Ziegler-Nichols,
Internal Model Control, and Lambda tuning among others. Once the auto-
tune test has been completed, the user can also switch between different algo-
rithms instantaneously to compare the different outcomes based on the differ-
ent algorithms. The auto-tuner can, in most cases, perform the step tests, make
the measurements and solve the calculations in less time than that which is
required to manually perform a step test, thereby saving a considerable
amount of time during the start-up of a process or during the general tuning
of process loops.
Local Override
In use, the auto-tuner will take control of the loop being tuned and put the
loop into LOCAL OVERRIDE. In Figure 10-1, LOCAL OVERRIDE is dis-
played on the faceplate as LO. This will let the board operator know that
another program is controlling the output. During this period, the auto-tuner
has control of the control valve (or another final control element) and it will
make adjustments according to how it has been configured. This means that
the auto-tuner will move the final control element as it sees fit for the purpose
of measuring the process parameters to obtain the desired tuning results,
instead of maintaining process control during normal operations.
The Auto-tuner Application 203
Figure 10-1. Faceplate Showing the Control Loop in LOCAL OVERRIDE (LO)
204 Tuning of Industrial Control Systems, Third Edition
Step Size
The step size can be selected depending on several variables of the loop being
tuned, the most important of which is process gain. Be aware that by perform-
ing a step test, you are causing a disturbance in a running process, and the
goal is to keep the disturbance to a minimum. For loops with a high process
gain, a small step size should generally be used. A step size of 3% of scale or
less can be chosen if there is very little process noise in the loop. On some
loops, a step size as small as 0.5% of scale can produce good results. For nois-
ier loops or loops with a low process gain, a larger step size should be
selected. If you are unsure of the process gain, start with a smaller step size
and retest with a larger step size if necessary. Figure 10-2 shows the auto-
tuner application tuning page. You can see that the step size parameter (in the
lower left corner) is selectable by dropping down the step size selection box.
One method that will aid in selecting an initial step size is to examine the
plant historian. Look at the historical data that was recorded during recent
process upsets and see whether small output responses were needed to rectify
a large process upset. If so, you can assume that the process gain is large and
you should start with a small step size. On the other hand, if you notice that
The Auto-tuner Application 205
10-2. Applications
The first requirement in tuning any process is that the loop must be stable. If
the loop is in automatic mode and the process variable is stable, you should
have no problem with auto-tuning the process while it is in automatic mode.
If the historical trend shows that the process variable is stable overall but has
some significant process noise, you may still be able to tune the loop using a
larger step size.
If stability cannot be achieved in automatic mode, you should switch the loop
to manual mode. Remember to monitor the process variable any time that the
loop is not under automatic control. If after switching the loop to manual
mode you still observe variability in the loop, then there is another loop acting
on the loop under test. You must find the loop that is interacting with the loop
under test and stabilize that loop first to obtain a steady process variable in
preparation for the tuning procedures. Keep in mind that other process prob-
lems, such as feed composition changes or equipment malfunctions, may also
cause interactions with the loop under test. Interacting loops may be easy to
locate if they are cyclic or recurring, but they may be harder to find if they are
intermittent and interact with the loop under test less frequently. The facility
process engineer may be able to offer helpful advice when it comes to tracking
down interacting loops.
206 Tuning of Industrial Control Systems, Third Edition
Not all loops can or should be tuned using the auto-tuner. In some cases, the
auto-tuner may not be able to achieve a sufficient response from the process
variable to provide accurate results. On some slower loops, the auto-tuner
may seem to hang up and not move the final control element as required to
generate oscillations in the process. When this happens, you have the option
to either wait and see whether the auto-tuner will switch states or to abort the
procedure and tune the loop using an alternate method.
In all cases, you should closely monitor the reaction of any loop in which the
tuning parameters have been modified. If the response of the loop is not
acceptable, you can retune the loop, or if necessary, revert to the original tun-
ing values until the process becomes stable again. We will discuss the applica-
tion of the auto-tuner later in this chapter.
The auto-tuner is just one tool that the engineer or technician has at his or her
disposal to help maintain the most efficient control of the process. A control
loop with a tight control response (when required) can save the facility unnec-
essary expense compared to a control loop that has high variability. For
instance, if you take the example of a process heater in which the fuel gas flow
is controlled by the heater outlet temperature, you can see where a steady con-
trol loop would cost less in expensive fuel gas when compared to a poorly
tuned loop in which the outlet temperature falls a few degrees and the control
valve opens a significant amount allowing a surge of fuel gas into the heater
to recover the few degrees of lost heat. Such a loop can be tuned using any
method, but if the time it takes to complete the tuning can be shortened by
using the auto-tuner, then the technician or engineer can move on to the next
loop sooner, thereby increasing profitability for the company.
Before attempting tuning or auto-tuning, you should ensure that all parts of
the process loop are working properly. Tuning will not rectify inherent prob-
lems within a loop, such as a valve that has hysteresis or stiction and is not
transitioning properly from opening to closing or from closing to opening.
Examine the location of the sensing element (sensor) in relation to the final
control element. If there is a large amount of piping in between the two, then
you might expect a long dead time in the process loop. Likewise, if the con-
trolled variable is on one product stream and the manipulated variable is on
another product stream, then it may take longer for the response to be noticed.
The Auto-tuner Application 207
The auto-tuner has special settings for slow time constant loops. These slow
time constant loops will usually take longer to auto-tune than standard loops.
They should be watched closely and the tuning should be aborted if the pro-
cess does not respond within a time period that will not interfere with normal
process operations.
The auto-tuner works well with fast loops. A typical flow loop usually has a
small time constant. Look back at Figure 10-2 and notice how quickly the step
test takes place. When this is the case, the auto-tuner can almost always finish
the testing and report the results faster than the process control engineer can
manually perform the step test. A loop that has a small time constant will also
settle back to normal operating conditions very quickly after the step test dis-
turbances have been injected. Most pressure loops are very fast by nature and
are easily tunable using the auto-tuner application. Some fast-acting level
loops are also good candidates for auto-tuning. Tight level control is easily
achievable using the auto-tuner on fast level loops. Loops with a small time
constant provide the most benefit from using the auto-tuner application
because of the relatively short amount of time it takes to complete the tuning
process using the auto-tuner.
There may be some loops in the process that would respond better if they
were not tightly tuned. For instance, the liquid level in a large tower may be
tuned less tightly to allow the level to fluctuate within reasonable limits if this
fluctuation does not cause disturbances in other loops (see Figure 10-3). If the
level control loop in the tower has a control valve that controls the level, and
the product leaving the tower is the feed to another unit, it may be more desir-
able to have the feed flow rate to the downstream unit remain more uniform
while allowing the level to fluctuate a bit more. In this case, the technician
could use the auto-tuner to achieve tight level control and then modify the
suggested tuning parameters to deaden the process response time a small
amount, or simply choose a more conservative tuning method, such as
Lambda tuning.
208 Tuning of Industrial Control Systems, Third Edition
Loops with a very small process gain should be monitored closely during the
auto-tune process. See Figure 10-4 for an example of a loop with a small pro-
cess gain. Notice that a step size of 12 percent of span was chosen for tuning
this particular loop. Also notice on the graph that the process variable only
moved a very small amount given the very large step size that was chosen.
Process Noise
Loops with a high level of process noise should also be monitored closely
during the tuning procedure. At times, the auto-tuner may actually mistake
process noise for a movement of the process variable PV in an incorrect
direction. In these cases, the auto-tuner may abort the tuning procedure,
forcing you to start over again. If possible, PV filtering may be used to limit
the amount of process noise seen by the auto-tuner. Remember that the auto-
tuner cannot distinguish between process noise and an actual process reading.
Use the minimum amount of filtering required to limit the process noise and
dont forget to set the filter back to normal (if required) after the tuning
process is completed.
In all tuning procedures the loop (or loops) should be evaluated closely after
any change in tuning parameters is made. Watch for the loops response to
load changes, upsets, and set point adjustments. If the loop does not perform
satisfactorily, you may try retuning the loop, or you may want to reinstall the
original tuning parameters. The auto-tuner has a button that will allow you to
revert to the original tuning parameters easily, should it be required.
The auto-tuner package has many features and settings that can make the job
of tuning an easy one for the process control technician or engineer. Many
control system manufacturers include auto-tune software with their control
system. Differences in available features may vary greatly between different
manufacturers. You must read the manufacturers documentation and
become familiar with the features of the software that you are using before
attempting tuning using the auto-tune software.
If an auto-tune software package is not included with the control system soft-
ware, it may have to be purchased separately, or a license may have to be pur-
chased separately from the control system license to allow the software to
operate properly.
Desired Response
Most auto-tuner software has a feature that will allow the user to choose cer-
tain options in the tuning such as Normal, Fast, or Slow to obtain a particular
desired response from the tuning. The Fast option allows the user to try a
210 Tuning of Industrial Control Systems, Third Edition
more aggressive approach for loops that need to return to set point quickly,
with overshoot being of a minimal concern. The Slow option reduces the
amount of overshoot experienced at the cost of PV being away from set point
longer. The Normal option is a trade-off between a quick return to set point
and a minimal amount of overshoot. Figure 10-5 illustrates the DESIRED
RESPONSE parameter in the highlighted drop-down box. As you can see, in
this case a NORMAL response has been selected.
Tuning Methods
Some expanded features of the software will allow you to choose which tun-
ing method is employed to enhance the performance of the loop. You may
choose Ziegler-Nichols, Lambda, or Internal Model Control for more desirable
tuning results. One of the best things about the auto-tuner is that once you
have performed the step test and recorded the process dynamics of the loop,
you do not have to retest each time you wish to switch between tuning meth-
ods. By simply selecting the new tuning method, you will instantly have the
new calculation results and will be able to quickly compare them to one
another with just a few clicks of the mouse.
The Auto-tuner Application 211
As you can see, the recommended setting is represented by a dot with REC-
OMMENDED beside it on the Tuning for Robustness graph. The second dot
depicts the simulated tuning value. It can be placed anywhere in the Tuning
for Robustness graph by simply clicking the mouse in the shaded area. The
auto-tuner will instantly recalculate the predicted response to the simulated
tuning values and plot the results in graphical form in the Simulation
Response area of the auto-tuners Simulate tab.
212 Tuning of Industrial Control Systems, Third Edition
10. Monitor the newly installed tuning parameters and control loop response.
10-4. Summary
The auto-tuner application does not replace the knowledge necessary to per-
form tuning procedures. The auto-tuner can drastically reduce the amount of
time required for manual tuning step testing and the required mathematical
calculations, but it must be properly configured before beginning any tuning
procedures.
The application will take control of the final control element and cause it to
move to generate a measurable response in the process variable. Constant
communication with the process operators is a must during the period that
the control loop is in LOCAL OVERRIDE.
The Auto-tuner Application 213
The user can select a certain desired response before testing. Once the process
dynamics have been measured, the user can switch between several of the
most common tuning methods available. The auto-tuner application also has
the ability to project expected responses to process disturbances based on
input from the user using the simulation feature that is included with the
auto-tuner.
Review Questions
10-1. Briefly describe LOCAL OVERRIDE and explain its effects on the con-
trol loop.
10-2. While selecting a step size for auto-tuning a control loop, you notice that
the process historian shows that the last upset took large movements of
the final control element to bring the process variable back to set point.
What can you assume about this loop and what step size should you
begin with?
10-4. What types of loops does the auto-tuner work best on?
10-5. Explain why the user must maintain constant contact with the board
operator while performing tuning procedures.
10-6. Describe the Simulate feature and how it outputs the results of simu-
lated tuning.
Appendix A
Suggested Reading and
Study Materials
Blevins, T. L., McMillan, G. K., Wojsznis, W. K., and Brown, M. W., Advanced
Control Unleashed: Plant Performance Management for Optimum Benefit, ISA,
Research Triangle Park, NC.
McMillan, G. K., and Toarmina, C. M., Advanced Temperature Control, 2nd edi-
tion, ISA, Research Triangle Park, NC.
McMillan, G. K., Good Tuning, A Pocket Guide, 3rd edition, ISA, Research Trian-
gle Park, NC
Shinskey, F. G., Process Control Systems, 3rd ed., McGraw-Hill, New York, NY,
1989.
Smith, C. A., and Corripio, A. B., Principles and Practice of Automatic Process
Control, 3rd edition, Wiley, New York, NY, 2006. (Note: The material on
Dynamic Matrix Control is in the 2nd edition, 1997).
215
Appendix B
Answers to Study
Questions
Chapter 1
1-2. The two process characteristics to be considered when tuning the con-
troller are the process sensitivity or gain and its rate of response
(Section 1-1).
1-4. The fourth element of the feedback loop is the process (Section 1-2).
1-5. The most important characteristic of a feedback control loop is its action,
direct or reverse, chosen so that there will be a net reverse action around
the loop; that is, when the process variable changes the controller will
take action to move it in the opposite direction (Section 1-2).
1-6. The fail position of the cooling water valve must be open so that coolant
is not lost on loss of power; the controller signal then closes the valve, so
the controller action must be reverse. In other words, when the tempera-
217
218 Tuning of Industrial Control Systems, Third Edition
ture increases the controller output decreases, opening the valve to sup-
ply a higher-cooling water flow.
1-7. The reactants control valve must fail closed so that the reactor does not
overflow on loss of power. The controller action must then be reverse;
that is, increasing level decreases the controller output to close the valve
and reduce the flow of reactants into the reactor.
1-8. The controller must have direct action; that is, increasing caustic compo-
sition must increase the controller output to increase the flow of dilution
water and decrease the composition of the caustic product.
Chapter 2
2-4. The cause is that the controller has the incorrect action. To correct,
change the controller action.
2-5. The problem is that the controller gain is too high (or the integral time is
too low). To correct, decrease the controller gain by at least half. If this
does not work, adjust the integral time using the period of the oscilla-
tions as a guide; that is, set the integral time to be of the order of magni-
tude of the period of the oscillations.
Answers to Study Questions 219
2-6. The integral and derivative times in Table 2-1 are related to the period of
oscillation of the loop because the period is an indication of the speed of
response of the loop.
2-7. To reduce the variability of the controller output, reduce the controller
gain.
Chapter 3
3-1. The procedure for the open-loop step test is as follows (Section 3-1):
c. Record on the same trend the controller output and the process
variable until the PV reaches a new steady value.
d. Analyze the PV response to obtain the gain, dead time, and time
constant.
3-2. The parameters of the SLPDT model are the gain, the dead time, and the
time constant (Section 3-2). The gain is an indication of the sensitivity of
the process variable to the controller output; the dead time is a measure
of how long it takes the PV to start responding to the change in control-
ler output; and the time constant is a measure of how long it takes the
process to respond to the action of the controller.
Note that 1 min is subtracted from the time readings because the step
change in controller output is applied at 1 min.
3-8. At the base conditions the total flow is (Sections 3-2, 3-3):
10 + 40 = 50 gpm
Answers to Study Questions 221
For a tank volume of 5,000 gal, the time constants are (Section 3-3,
Equation 3):
After the change in flow, they are 9.8 and 98 min, respectively.
These answers illustrate the effect of the process flow on the gain and
time constants of the process model due to a simple and common
nonlinearity.
Chapter 4
4-1. If the process gain doubles due to nonlinear behavior, the controller
gain must be halved to maintain the same loop gain and controller per-
formance (Section 4-1).
4-2. The uncontrollability is the ratio of the dead time to the time constant.
The actual magnitude of the time constant and the dead time determine
how fast the loop can respond, but not its controllability (Section 4-1).
4-3. The proportional gain from the ultimate gain is Kcu/2 and from the
uncontrollability parameter is 1/KPu. For them to be the same, Kcu = 2/
KPu. The PID integral time from the ultimate period is Tu/2 and from
the dead time is 2t0, so for them to be the same, Tu = 4t0 (Section 4-1).
4-4. The sensitivity is the measure of how much the process variable changes
when the controller output changes; that is, the gain, so Process A is the
least sensitive and Process C the most. The speed of response is deter-
222 Tuning of Industrial Control Systems, Third Edition
mined by the process time constant, so Process B is the slowest and Pro-
cess A the fastest. The uncontrollability is the ratio of the dead time to
the time constant, so Process C is the least controllable and Processes A
and B are equally controllable (Section 4-1).
4-5. Using the tuning strategy proposed in the chapter, the PID tuning
parameters are (Section 4-1):
Process A: Kc = (1.2/0.5)(2.0/0.2) = 24
TI = 2(0.2) = 0.4 min
TD = 0.2/2 = 0.1 min
Process B: Kc = (1.2/2.0)(30/3.0) = 6.0
TI = 2(3.0) = 6.0 min
TD = 3.0/2 = 1.5 min
Process C: Kc = (1.2/4.0)(5.0/3.0) = 0.5
TI = 5.0 min (use IMC rule, 2(3) > 5)
TD = 3.0/2 = 1.5 min
4-6. The controller would be configured to act on the process variable to pre-
vent large changes in controller output when the set point is changed
and the controller gain is high. When configured in this manner, the
output ramps to the new required value at the rate controlled by the
integral time when the set point is changed (Section 4-1).
4-8. Inverse response is the case where the process variable initially moves
in the opposite direction of its eventual direction of change. It is detri-
mental to the performance of a feedback controller because it causes the
controller to initially move the output in the wrong direction
(Section 4-4).
Answers to Study Questions 223
Chapter 5
5-1. Tight level control is indicated when the level has significant effect on
the process operation, as in a natural-circulation evaporator or reboiler.
Averaging level control is to be used when it is necessary to smooth out
sudden variations in flow, such as in a surge tank receiving discharge
from batch operations to feed a continuous process. Tight level control
is the one that requires the level to be kept at or very near its set point
(Section 5-3).
5-3. For tight level control a proportional controller with a high gain, usually
greater than 10, should be used. When the lag of the control valve is sig-
nificant, a proportional-derivative controller could be used. If a propor-
tional-integral controller is used, the integral time should be long, on the
order of one hour or longer (Section 5-3).
5-4. For averaging level control a proportional controller with a gain of 1.0
should be used, because this provides maximum smoothing of varia-
tions in flow while still preventing the level from overflowing or run-
ning dry (Section 5-3).
5-5. When a PI controller is used for averaging level control, the integral
time should be long, on the order of one hour or longer. At some values
of the gain, an increase in gain would decrease oscillations in the flow
and the level (Section 5-3).
5-7. PID controllers are commonly used for temperature control so that the
derivative mode compensates for the lag of the temperature sensor,
which is usually significant (Section 5-4).
5-8. The major difficulty with the control of composition is the dead time
introduced by sampling and by the analysis (Section 5-5).
224 Tuning of Industrial Control Systems, Third Edition
Chapter 6
6-2. The derivative kick is a pulse on the controller output that takes place
at the next sample after the set point is changed and lasts for one sam-
ple. It can be prevented by having the derivative term act on the process
variable instead of on the error (deviation) (Section 6-1).
6-3. Proportional kick is a step change in controller output right after a set
point change; it can be eliminated by having the proportional term act
on the process variable instead of on the error, so that the operator can
apply large changes in set point without danger of upsetting the pro-
cess. When proportional kick is avoided, the process variable
approaches the set point slowly after it is changed, at a rate determined
by the integral time. The proportional kick must not be avoided when-
ever it is necessary to have the process variable follow set point changes
quickly, as is the case in the secondary controller of a cascade system
(Section 6-1).
6-4. The nonlinear gain allows the proportional controller gain to be smaller
than unity when the error is near zero, which is equivalent to having a
larger tank in an averaging level control situation. To have a gain of 0.2
(400% PB) at zero error, the nonlinear gain must be (Equation 6-1):
6-5. Using the formulas of Table 6-2, with q = 0 (for maximum gain) and the
following parameters (Section 6-2):
Answers to Study Questions 225
6-6. If the algorithm has dead time compensation, the gain can be higher
because it does not have to be adjusted for dead time. This does not
affect case (d) because the dead time is less than one sample, and there-
fore, no dead time compensation is necessary. From Equation 6-5 and
Table 6-3 (Section 6-4):
6-7. The basic idea of the Smith Predictor is to bypass the process dead time
to make the loop more controllable. This is accomplished with an inter-
nal model of the process responding to the manipulated variable in par-
allel with the process. The basic disadvantage is that a complete process
model is required but it is not used to tune the controller, creating too
many adjustable parameters (Section 6-4).
Chapter 7
7-1. Cascade control (1) takes care of disturbances into the secondary loop
reducing their effect on the controlled variable, (2) makes the primary
loop more controllable by speeding up the secondary loop and (3) han-
dles the nonlinearities in the inner loop, where they have less effect on
controllability (Section 7-1).
226 Tuning of Industrial Control Systems, Third Edition
7-2. For cascade control to improve the control performance, the secondary
loop must be faster than the primary loop. The sensor of the secondary
loop must be reliable and fast, although it does not have to be accurate
(Section 7-1).
7-3. The primary controller in a cascade control system has the same require-
ments as the controller in a simple feedback control loop, thus the tun-
ing and mode selection of the primary controller are no different from
those for a single controller (Section 7-2).
7-5. The controllers in a cascade system must be tuned from the inside out,
because each secondary controller forms part of the process controlled
by the primary around it (Section 7-3).
7-9. Integral windup can occur in cascade control when the operating range
of the secondary variable is narrower than the transmitter range. To pre-
Answers to Study Questions 227
vent it, the secondary measurement can be passed to the integral feed-
back of the primary; in such a scheme the primary always takes action
based on the current measurement, not on its set point (Section 7-4).
Chapter 8
8-1. A feedback controller acts on the deviation of the process variable from
the set point. Thus, if there is no deviation, there is no control action. In
theory, perfect control is possible with feedforward control, but it
requires perfect process modeling and compensation (Section 8-1).
8-2. To be used by itself, feedforward control requires that all the distur-
bances be measured and accurate models be developed of how the dis-
turbances and the manipulated variable affect the controlled variable
(Section 8-1).
Feedforward with feedback trim has the advantage that only the major
disturbances have to be measured and compensation does not have to
be exact, because the integral action of the feedback controller takes care
of the minor disturbances and the model inaccuracies.
8-3. Ratio control consists of maintaining constant the ratio of two process
flows by manipulating one of them. It is the simplest form of feedfor-
ward control (Section 8-1).
The response of a lead-lag unit to a ramp is a ramp that leads the input
ramp by the difference between the lead and the lag time constants, or
lags it by the difference between the lag and the lead time constants.
Therefore, a lag of 1.5 minutes and a lead of 3.0 minutes (Section 8-3).
Dead time compensation can be used only when the feedforward action
is to be delayed and a computer or microprocessor device is available to
implement it. It should be used only when a simple lag cannot
effectively delay the feedforward compensation.
8-7. Design of feedforward controller for the process furnace (Section 8-4):
Solve for the manipulated variable and substitute the control objective:
Design formula:
7. Lead-lag units must be installed on the process flow and inlet tem-
peratures, but not on the supplementary fuel gas flow, because its
dynamic effect should match that of the main fuel gas flow
(Section 8-4).
Figure B-8-1.
TT TT Toset FT FT
Ti To SP W
Lead-Lag Lead-Lag
TC 2
1
OP +
- Fs
Adder
Toset - Ti
Multiplier
+
-
Adder
Fset
230 Tuning of Industrial Control Systems, Third Edition
Chapter 9
9-1. Loop interaction takes place when the controller output of each loop
affects the process variable of the other loop. The effect is that the gain
and the dynamic response of each loop change when the auto/manual
state or tuning of the other loops changes (Section 9-1).
When loop interaction is present, we can (1) pair the loops in the way
that minimizes the effect of interaction and (2) design a control scheme
that decouples the loops.
9-2. Open-loop gain of a loop is the change in its process variable divided by
the change in its controller output when all other loops are opened (in
Manual) (Section 9-2).
Closed-loop gain is the gain of a loop when all other loops are closed (in
Automatic) and have integral mode.
Relative gain (interaction measure) for a loop is the ratio of its open-loop
gain to its closedloop gain.
9-3. To minimize interaction for a loop, the relative gain for that loop must
be as close to unity as possible. Thus, the loops must be paired to keep
the relative gains close to unity, which in a system with more than two
control objectives may require ranking the objectives (Section 9-2).
The relative gains are easy to determine because they involve only a
steady-state model of the process, which is usually available at design
time.
The main shortcoming of the relative gain is that it does not take into
account the dynamic response of the loops.
9-4. When all four relative gains are 0.5, the effect of interaction is the same
for both pairing options. The gain of each loop will double when the
other loop is switched to Automatic. The interaction is positive; that is,
the loops help each other (Section 9-2).
9-5. When the effect of interaction with other loops is in the same direction
as the direct effect for that loop, the interaction is positive; if the interac-
tion and direct effects are in opposite direction, the interaction is nega-
tive. For positive interaction, the relative gain is positive and less than
Answers to Study Questions 231
unity, while for negative interaction the relative gain is either negative
or greater than unity (Section 9-2).
The top composition must be paired to the reflux and the bottoms
composition to the steam to minimize the effect of interaction.
9-7. Let H be the flow of hot water in gpm, let C be the flow of cold water in
gpm, let F be the total flow in gpm and let T be the shower temperature
in F. The mass and energy balances on the shower, neglecting varia-
tions in density and specific heat, give the following formulas:
These are the same formulas as for the blender of Example 9-2. The
relative gains are therefore:
| Hot Cold
F | H/F C/F
T | C/F H/F
Since the cold water flow is the higher flow, use it to control the flow
and use the hot water flow to control the temperature. The relative gain
for this pairing is (Section 9-2):
9-8. As in the second part of Example 9-4, we can use a ratio controller to
maintain a constant temperature when the flow changes. We would
then ratio the hot water flow (smaller) to the cold water flow (larger)
and manipulate the cold water flow to control the total flow. The design
ratio is 0.5 gpm of hot water per gpm of cold water (Section 9-3).
Chapter 10
10-1. LOCAL OVERRIDE is when another program takes control of the out-
put and moves it to measure process parameters to obtain tuning
results.
10-2. The loop has a small process gain and you should start with a larger
step size.
10-5. If the process starts to become unstable, the board operator will need to
retake control of the loop immediately.
10-6. The Simulate feature allows the user to simulate loop response for dif-
ferent tuning settings. The simulated response is presented in the form
of a chart that displays the projected amount of overshoot versus the
time required to reach set point again.
INDEX
batch process 51 66 87
bias 82 101102 118
224
blending tank 42 44 49
86
capacitance 4145 72
cascade control 78 18 80
100 121125 127129
131 133 135138
140 143 225226
cascade dontrol 130 135
characteristics
control loop 21
control valve 5051
derivative mode 16
feedback control 144
feedback control loops 19
feedforward control 144
loop interaction 173
PI controllers 86
process 12 29 41
74 217
clamp limit 138
clamped 137
closed-loop gain 175178 199 230
closed-loop time constant 106
coarse tuning 67
compensation for dead time 113 156
composition control 54 77 180
190 193196
computer cascade control 141
conductance 4146
conductance, valve 4344 46 50
54
derivative (Cont.)
kick 97 100 118
224
mode 1416 21 25
27 58 61
6667 7778 82
97 110 112
126 128 140
218 223 226
time 1520 24 2830
5860 68 82
9799 106 127128
224
unit 98100 127128
derivative-on-PV 100
desired response 209210
digital controller 136
direct action 67 40 173
194 218
direct material balance control 184185
distillation column 45 48 129130
170 181184 199
distributed control systems 5 18 96
148 160
disturbance 12 5 7
19 21 2324
31 34 6267
6972 78 82
92 100 107
112113 122 124
126 133 136
144147 149154 156
161 164165 171
186 203205 207
disturbance (Cont.)
213 222 225
227 229
dynamic compensation 116 146 151152
156 159 163
165 187189
dynamic gain limit 98 118 224
dynamic interaction 185 199
gain 3 1112 16
18 20 22
2426 30 34
3637 3940 50
53 5758 6061
64 74 76
80 82 87
92 101 105
107 110 113
126 128 132
148 151153 156157
162 177 179
181 184185 190
194196 199 217
219 223 225
230
closed-loop 175178 199 230
nonlinear 101102 118 224
open-loop 174178 180 183184
199200 230
relative 175180 183 185186
189 193196 198200
230231
steady-state 158 178 180181
195
variation 53
gap 7071 113 137
139
gas surge tank 42 44 79
graceful degradation 160
IMC 60 6264 74
80 92 132
222
independent variables 198
integral controller 7980 92 108109
223
integral mode 1314 21 25
51 58 66
7071 7879 82
85 112 124
126 136 139140
lag 41 153
lagniappe 123
lead 153
lead-lag compensation 163
lead-lag unit 144 151153 155156
163164 166 187
227228
level control 72 8182 8487
92 96 118
125 182 207
223224
linear feedforward controllers 149 152
liquid storage tank 4244
local override 202203 205 212213
local set point 18
loop interaction 170 173174 186
198199 230
manipulated variable 5 22 60
68 7071 78
88 149 151152
160 162 166
174 206 217
225 227228
manual output 34
master controller 7 138
material balance control 184185
measured disturbance 144 147148 158161
163 166 197
227228
negative feedback 6
negative interaction 173 179 181
185186 188 194195
199 231
noise 209
nonlinear feedforward compensation 158
nonlinear gain 101102 118 224
nonlinearity 58 148 221
output (Cont.)
116 118 122
125 128 131
133 135137 139
144 146 148150
152153 155159 162163
169171 173174 176179
181 186 188
190 194 197199
202 204 218219
221222 224 229230
overshoot 63 6971 100
125 127 132
136137 139 173
188 194195 210211
222 226 232
performance (Cont.)
158 165 169
173174 181 185
188 210 221222
226
pH control 101
PI controller 2526 30 58
6163 74 79
8586 92 97
105106 109110 115116
132 192 223
PID 8
PID algorithm 96
PID controller 1112 17 21
2427 30 58
6164 72 76
9293 106 110
114 132 223
pneumatic 96
positive interaction 173 179 185
195 230
pressure control 8081 86 92
112 125 136
primary controller 122
problematic loops 206
process dead time 37 48 59
76 106 109
111 113 225
process gain 33 50 5253
55 5759 65
68 109 122
204 208 220221
232
process noise 209
process nonlinearity 33 51 74
process time constant 37 41 60
86 91 106
108 111 222
process variable 13 58 1219
25 3436 41
46 64 6667
71 76 78
82 8889 9192
9598 100 102
106 110 112
115 118 121
125128 131 136
139140 144145 149
151152 158159 166
169174 176 178179
181 187188 192193
197199 205206 208209
212213 217 219
221222 224 226227
230
processing frequency 111
programmable logic controllers (PLC) 96
proportional band 1718 78 82
proportional controller 2123 2526 29
78 8586 90
92 102 223
proportional gain 12 17 19
22 2728 58
65 68 78
80 82 8586
97 101 118
125126 133 218
221 226
QDR response 23 25 58
74 133
QDR tuning 24 26 28
6163 132
quarter-decay ratio (QDR) response 58
rate time 14 58
ratio control 78 143144 148149
166 179 227
reactor 9 41 54
7073 122123 126
reactor (Cont.)
129 131133 135
137 139 218
relative gain 175180 183 185186
189 193196 198200
230231
matrix 174 178 185
reset feedback 138139 226
reset time 13 58
reset windup 51 57 6971
74 76 121
128 130 136
138 140141 222
resistance 4243 45 86
resistance temperature device (RTD) 89
reverse 6
reverse action 6 19 217
running away 19
stability (Cont.)
165 188 192
205
static compensation 163164 187
static compensator 152
static friction 80
steady-state 52 55 71
78 158 160
165 172 174175
177178 228
steady-state gain 158 178 180181
195
steam heater 115116 148 160
step size 204
step test 33 3536 3941
48 5354 57
68 103 106
131132 202 204
207 210 212
219
stiction 206
subcritical 46
swell 72
temperature control 4 6 19
25 35 70
72 87 9091
108110 117 123124
129 139 160
165 167 223
temperature-to-flow control scheme 128
three-mode controller 18
tight control 68 81 83
87 92 206
time constant 3334 3746 48
50 5355 5760
6264 68 74
76 7880 82
8586 89 91
93 97 103106
108 111113 115
119 128129 132
135136 144 151
153 156 189
191 194 207
219223 227228 232
time delay 4647 91 145
228
transducer 5
transfer function 113
transportation lag 36 40 4648
163
tuning 1
tuning for robustness 211
tuning methods 210
tuning parameter 23 1214 17
2429 6163 6768
76 104 107
109 114 116
119 126 193
198 206207 209
212 222
two-mode controller 18
two-point method 181
windup 51 57 6971
74 76 121
128 130 136
138141 222 226