0% found this document useful (0 votes)
35 views20 pagesPid
Uploaded by
Shubham KumarCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
35 views20 pagesPid
Uploaded by
Shubham KumarCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 20
CONTROLLER PRINCIPLES | 457
9.5 CONTINUOUS CONTROLLER MODES
‘The most common controller action used in process control is one or a combination of con-
tinuous controller modes. In these modes, the output of the controller changes smoothly in
response to the error or rate of change of error. These modes are an extension of the dis
continuous types discussed in the previous section.
9.5.1 Proportional Control Mode
‘The two-position mode had the controller output of either 100% or 0%, depending on the
crror being greater or less than the neutral zone. In multiple-step mode8, more divisions
of controller outputs versus error are developed. The natural extension of this concept is
the proportional mode, where a smooth, linear relationship exists between the controller
‘output and the error. Thus, over some range of errors about the setpoint, each value of er-
ror has a unique value of controller output in one-to-one correspondence. The range of
error f0 cover the 0% to 100% controller output is called the proportional band, because
the one-to-one correspondence exists only for errors in this range. This mode can be ex-
pressed by
P= Key +b 0.4)
where Kp = proportional gain between error and controller output (% per %)
Po = controller output with no error (%)
Direct and Reverse Action Recall that the ertor in Equation (9.14) is expressed
using the difference between setpoint and the measurement, ~ b. This means that as the
‘measured valve increases above the setpoint, the ertor will be negative andthe output will
decrease. Thats, the term Kp e, will subtract from py. Thus, Equation (9.14) represents re-
verse ction. Direct action would be provided by patting @ negative sign infront of the eor-
rection tem.
A plot of the proportional mode output versus error for Equation (9.14) is shown in
Figure 9.11. In this case, pg has been set to 50% and two different gains have been used.
‘Note that the proportional band is dependent on the gain. A high gain means large response
{0 an error, bat also a narrow error band within which the output is not saturated.
In general, the proportional band is defined by the equation
_ 0
Boe
(9.15)
Let us summarize the characteristics of the proportional mode and Equation (9.14).
1. Ifthe error is zero, the oupat is a constant equal 0 Pe
2. I there is error, for every 1% of error, a comrection of Kp percent is added to or
subiracted from py, depending. on the sign of the eror.
3. There is a band of error about zero of magnitude PB within which the output is
not saturated at 0% oF 100%. 458 | CHAPTER 8
Saturation
100
Controller output (24)
8
FIGURE 9.11
‘The proportionel band of a proportional controller depends on the inverse of the gain.
100
Conroe output (4)
Error (%)
FIGURE 9.12
An offset error must occur i @ proportional controller requires a new zero-error output fol
lowing a load chengs
Offset An important characteristic of the proportional control mode is that it pro-
duces a permanent residual error in the operating point ofthe controlled variable when a
change in load occurs. This error is referred to as offset. It can be minimized by a larger con-
‘stant, Kp, which also reduces the proportional band. To see how offset occurs, consider asys-
tem under nominal load withthe controller at 50% andthe error zero, 2s shown in Figure 9.12 CONTROLLER PRINCIPLES | 459
I transient error occurs, the system responds by changing controller output in correspon-
dence with the transient to effect a retum-to-zero error. Suppose, however, a load change oc-
‘curs that requires a permanent change in controller output to produce the zero-error state,
Because a one-to-one correspondence exists between controller output and exo, itis clear that
‘anew, zero-crror controller output can never be achieved. Instead, the system produces 2 small
permanent offset in reaching 2 compromise position of controller output under new loads.
EXAMPLE Consider the proportional-mode level-contto system of Figure 9.13. Value is linea, with
96 a flow seale factor of 10 m’/h per percent controller output. The controler output is nom-
inally 50% witha constant of Kp = 10% per %. A load change occurs witen flow through
valve B changes from 500 mh to 600 m'/b, Calculate the new controller output and off
set eror.
Solution
Certainly, valve A must move to a new postion of 600 m/h flow or the tank will empty.
This can be accomplished by a 60% new controller output because
on (222) com = 600 m'/h
as required. Because this isa proportional controller, we hve
P= Kre, + Po
with the nominal condition pp = S05 Thus
Pom W~ 50
oe 7 wo
ep = 1%
$02 1% offset error occurred because ofthe load change.
Consol vate
7 Ccontratr Le-—— Setpoint
FIGURE 9.13
Level-control system for Example 9.6. 460 | CHAPTER 9
Application The offset error limits use of the proportional mode to only a few
cases, particularly those where a manual reset ofthe operating point is possible to eliminate
offset. Proportional conttol generally is used in processes where large load changes are un
likely or with moderate to small process lag times. Thus, ifthe process lag time is small,
the proportional band can be made very sunall (large Kp), which reduces offset error.
Figure 9.11 shows that if Kp is made very large, the PB becomes very small, and the pro-
portional mode acts just like an ON/OFF mode. Remember that the ON/OFF mode exhib-
ited oscillations about the setpoint. From these statements itis clear that, for high gain, the
proportional mode causes oscillations of the exror.
9.5.2 Integral-Control Mode
‘The offset error of the proportional mode occurs because the controller cannot adapt to
changing external conditions—that is, changing loads, In other words, the zero-error out
put is @ fixed value. The integral mode eliminates this problem by allowing the controller
{o adapt to changing extemal conditions by changing the zero-error output.
‘The need for integral action shows up when its noted that even with proportional ac-
tion correction, the error does not go to zero in time. Suppose a system has some exror, ep,
‘nd the proportional mode provides a change in controller output, Kpe,, As we watch the
error in time, we note that the error may reduce, but it does nor go fo zero; in fact, it may
become constant. Integral action is needed.
Integral action is provided by summing the error over time, multiplying that sum by a
‘ean, and adding the result to the present controller output. You can see that ifthe error makes
random excursions above and below zero, the net sum will be zero, so the integral action will
not contribute, But if the error becomes positive or negative for an extended period of time,
‘the integral action will begin fo accumulate and make changes to the controller output,
In the mathematics of continous functions, such as error, summation is represented
bby integration, Therefore, this mode is represented by an integral equation
p(t) = Kifer dt + p(0) (9.16)
‘where p(0) is the controller output when the integral action starts. The gain K, expresses
how much controller output in percent is needed for every pervent-time accumulation
of error.
Another way of thinking of integral action is found by taking the derivative of Equa-
tion (9.16), In that case, we find a relation forthe rate at which the controller output changes,
dp
an 7 Kite (17)
‘This equation shows that when an error occurs, the controller begins to increase (or de-
crease) its outpot ata rate that depends upon the size of the error and the gain. Ifthe etror
js zero, the controller output is not changed. If there is positive error, the controller output
begins to ramp up at a rate determined by Equation (9.17), Figure 9.14 illustrates this for
‘wo different values of gain, CONTROLLER PRINCIPLES | 461
(0)
‘Li...
FIGURE 9.14
integral mocie controller action: (a) The rate of output change depends on error, and {b) an
ilustration of integral made output and ertor
Figure 9.14a shows how the rate of change of controller output depends upon the
value of error and the size of the gain, Figure 9.14b shows how the actual controller output
‘would lookif a constant error occurred. You can see how the controller output begins to ramp
up ata rate determined by the gain. In the case of gain K;, the output finally saturates at
100%, and no further action can occur (pethaps a control valve i fully open, for example)
Let us summarize the characteristics ofthe integral mode and Equation (9.16).
1, Ifthe error is zero, the output stays fixed at a value equal to what it was when the
ettor went to zero,
2. If the error is not zero, the output will begin to increase or decrease at a rate of
K, percenu/second for every 1% of excor. 462 | CHAPTER 8
‘Area Accumulation From calculus we learn that an integral determines the area
‘of the function being integrated, Thus, Equation (9.16) can be interpreted as providing @
‘controller output equal to the net area under the error-time curve multiplied by K,, We of-
ten say that the integral term accumulates error as a function of time, Thus, for every
1% ~ s of accumulated error-time area, the output will be Ky percent.
EXAMPLE
97
‘An integral controler is used for speed control with a setpoint of 12 rpm within a range of
10:0 15 rpm, The controller output is 22% intially. The constant Ky = ~0.15% controller
owipat per second per percentage eror. Ifthe speed jumps to 13.5 tpm, calculate the con-
twoller output after 2s fora constant e
Solution
We find ¢, from Equation (9.3):
AS%/s
but because ¢, is constant,
p= Kept + 700)
‘Afler 2 s, we have
p= (0.15)(30%)(2) + 22
3%
‘The integral gain, K,, is often represented by the inverse, which is called the integral rime,
‘or the reset action, T; = 1/K;. This is often expressed in minutes instead of seconds be-
‘cause this unit is more typical of many industrial process speeds.
‘The integral controller constant K, may be expressed in percentage change per
minute per pereentage error, whenever a typical process-control loop has characteristic re- CONTROLLER PRINCIPLES | 463
sponse time in minutes rather than seconds. Thus, an integral mode controller with reset ac-
‘ion at 5.7 minutes means that K; for our equations would be
1
Gy minj(6Os/miny
Ky = 292 x 10st
Applications Use of the integra! mode is shown by the flow control system in
Figure 9.9, except that we now assume a reverse-acting integral controller mode. Opera-
tion can be understood using Figure 9.15. A load change-induced error occurs at ¢ = 0,
Ener 0
‘Time
| = Process and controle ag
| conoteriag
Contre
output (2%)
wor 4
L L L
FIGURE 9.15
Iustration of integral mode output and error, showing the effect of process and contol lag 484 | CHAPTER 9
"The proper valve position under the new load to maintain the constant flow rate is shown
asa dashed line in the controller output graph of Figure 9.15. In the integral mode, the value
initially begins to change rapidly, as predicted by Equation (9.17). As the valve opens, the
error decreases and slows the valve opening rate as shown. The ultimate effect is that the
system drives the error to zero at a slowing controller rate. The effect of process and con-
‘tol system lag is shown as simple delays in the controller output change and in the error
reduetion when the controller action occurs. Ifthe process lags are too large, the error can
oscillate about zero or even be cyclic. Typically, the integral mode is not used alone, but
ccan be used for systems with smal process lags and correspondingly small capacities.
9.5.3 Derivative-Control Mode
Suppose you were in charge of controlling some variable, and at some time, f, your helpet
yelled out, “The error is zero, What action do you want to take’”” Well, it would seem per-
fecily rational to answer “None” because, after al, the error was zero. But suppose You
have a screen that shows the variation of error in time and that it looks like Figure 9.16.
‘You can clearly see that even though the ertor at fs ze10, itis changing in time and
‘will certainly not be zero in the following time. Therefore, some action should be taken even
though the error is zero! This scenario describes the nature and need for derivative action.
Derivation controller action responds to the rate at which the error is changing—
that is, the derivative of the ervor. Appropriately, the equation for this mode is given by
the expression
ent
20) = Bo
(9.18)
where the yn, Kp tells us by how much percent to change the controller output for every
percent-per-second rate of change of error. Derivative ation is not used alone because it
provides no output when the error is constant.
Derivative controller action is also called vate action and anticipatory control
FIGURE 9.16
The error can be zero but the rate of change very large. CONTROLLER PRINCIPLES | 465
Figure 9.17 illustrates how derivative action changes the controller output for vari-
‘ous rates of change of error, For this example, itis assumed that the controller output with
no error or rate of change of error is 50%. When the error changes very rapidly with a pos-
itive slope, the output jumps to a large value, and when the error is not changing, the out-
put returns to 50%, Finally, when the error is decreasing—that is, has a negative slope—the
‘utpat discontinuously changes to a lower value.
TThe derivative mode must be used with great care and usually with a small gain, be~
‘cause a rapid rate of change of error can cause very large, sudden changes of controller out-
put. Such an event can lead to instability
Let us summarize the characteristics of the derivative mode and Equation (9.18),
1. Ifthe error is zero, the mode provides no output.
2. Ifthe error is constant in time, the mode provides no output.
3. Ifthe error is changing in time, the mode contributes an output of Kp percent for
every 1Mb-per-second rate of change of error,
4, For direct action, a positive rate of change of error produces a positive derivative
mode output,
‘Time
Ccontrolter output
g
7 Time
FIGURE 9.17
Derivative mode controller action changes depending on the rate of error. 466 | CHAPTER 8
9.6 COMPOSITE CONTROL MODES
Itis common in the complex of industrial processes fo find control requirements that do not
fit the application norms of any of the previously consideted controller modes. It is both
possible and expedient to combine several basic modes, thereby gaining the advantages of
each mode. In some cases, an added advantage is that the modes tend to eliminate some tim-
itations they individually possess. We will consider only those combinations that are com-
monly used and discuss the merits of each mode.
9.6.1 Proportional-Integral Control (PI)
‘This is @ control mode that results from a combination ofthe proportional mode and the in-
tegral mode. The analytic expression for this control process is found froma series combi-
nation of Equations (9.14) and (9.16):
p= Kye, + KpKy [ ep dt + p4(0) 0.19)
where py(0) = integral term value at = 0 (Initia value)
‘The main advantage of this composite control mode is that the one-to-one corre-
spondence of the proportional mode is available and the integral mode eliminates the
inherent offset. Notice that the proportional gain, by design, also changes the net inte~
gration mode gain, but that the integration gain, through K;, can be independently ad
justed. Recall that the proportional mode offset occurred when a load change required
‘a new nominal controller output that could not be provided except by a fixed error from
the setpoint. In the present mode, the integral function provides the required new con-
troller output, thereby allowing the errarto be zero after a load change. The integral fea-
ture effectively provides a reset of the zero erfor output after a load change occurs. This
can be seen by the graphs of Figure 9.18. At time /, a load change occurs that produces
the error shown, Accommodation of the new load condition requires a new controller
output. We see that the controller output is provided through a sum of proportional plus
integral action that finally leaves the error at zero. The proportional part is obviously
just an image of the error,
Let us summatize the characteristics of the PI mode and Equation (9.19),
1. When the error is zero, the controller output i fixed at the value that the integral
term had when the error went to zero. This output is given by p/(0) in Equa-
tion (9.19) simply because we chose to define the time at which observation
stats as = 0.
If the error is not zero, the proportional term contibutes a correction, and the
imegral term begins to increase or decrease the accumulated value fintially,
‘p(0)} depending onthe sign ofthe error and the direct o reverse action,
‘The integral term cannot become negative. Thus, it will saturate at zero if the error
and action try to drive the area to a net negative valve. EXAMPLE
98
CONTROLLER PRINCIPLES | 467
”
Boo t
é fi}
efor ot '
i | '
1 1 1
it |
pm | it |
1 1 1
wor tt '
1
hot {
3 1 1 | Proportions!
& 1 1
3 1 ot Integral
3 1
3 4,
| '
FIGURE 9.18
Proportional.integral (PI) action showing the reset action of the integral contribution. This
‘example is for reverse action
Application As noted, this composite proportional-integral mode etiminates
the offset problem of proportional conrollers. It follows thatthe mode can be used in sys-
tems with frequent or large load changes. Because ofthe integration time, however, the
process must have relatively slow changes in load to prevent oscillations induced by the
integral overshoot. Another disadvantage ofthis system is that during start-up ofa batch
process, the integral action causes a considerable overshoot of the error and output be-
fore setting to the operation point. This is shown in Figure 9.19, where we see the pro-
portional band as a dashed band. The effect of the integral action can be viewed as a
shifting of the whole proportional band. The proportional band is defined as that positive
and negative error for which the output will be driven to O% and 100%. Therefore, the
presence of an integral accumulation changes the amount of error that will bring about
such saturation by the proportional term. In Figure 9.19, the output saturates whenever
the error exceeds the PB limits. The PB is constant, but its location is shifted as the inte~
gral term changes.
Given the error of Figure 9.20 (top), plot a graph of a proportional-integral controller out-
putas a function of time.
Kp =5,K,
10%
0s", and pr(0) 468 | CHAPTER 9
error
109 Tine
j
os SSC
Time
FIGURE 9.19
Overshoot and eycling often result when Pl mode control is used in startup of batch
processes. The dashed lines show the proportional band.
Solution
‘We find the solution by an application of
P= Kee, + KK, f edt + (0)
‘To find the controller output, we solve Equation (9.19) in time. The error can be expressed
in three time regions.
OS! (rbetween and 1s) CONTROLLER PRINCIPLES | 469
FIGURE 9.20
Solution for Example 9.8.
Error im)
Controter output (8)
‘The error rises from 0% to 19 in Is. Thus, itis given by ep
1sts3
For this time span, the error is constant and equal to 1%; therefore, i is given by e
123
For this time, the error is zero, e, = 0.
‘We now write out and solve Equation (9.19) for each of these time spans.
osrst
ym +5 frat +20
nases[E] f+
py = St + 250 + 20 470 | CHAPTER 9
‘This is plotted in Figure 9.20 (bottom) from 0 to 1 s. Notice the curvature because of the
squared term, Remember that only the integral term accumulates values. so in finding
the output at 1 5, the contribution of the proportional term, St, is not included. Therefore,
the starting value for the next time span is given by p,(1) = 2.5! + 20 = 22.
“The integral erm accumulation from Oto {forms the initia condition for this new equation,
pro 5+ 5[eh +225
p= S45(¢~ 1) +225
‘This function is plotted in Figure 9.20 from 1 to 3 s. At the end of this period, the integral
term has accumulated a value of p,(3) = 32.5%.
0
123
ps = 5[0] +5 [0 dr +325
Py = 325
Figure 9.20 (bottom) shows that the output will stay constant at 32.5% from 3 s. The sud-
den drop of 5% is due to the sudden change of error from 1% to 0% at ¢ = 3,
EXAMPLE
99
9.6.2 Proportional-Derivative Control Mode (PD)
A second combination of control modes has snany industrial applications. It involves the
serial (cascaded) use of the proportional and derivative modes. The analytic expression for
this mode is found from a combination of Equations (9.14) and (9.18):
(9.20)
where the terms are all defined in terms given by previous equations,
Its clear that this system cannot eliminate the offset of proportional controllers. It
can, however, handle fast process load changes as long as the load change offset error is ac-
ceptable. An example ofthe operation of this mode for a hypothetical load change is shown
in Figure 9.21. Note the effect of derivative action in moving the controller output in rela
tion tothe error rate change.
‘Suppose the error, Figure 9.222, is applied to a proportional-derivative controller with
Kp = 5, Ky = 0.55, and py = 20%, Draw a graph of the resulting controller output CONTROLLER PRINCIPLES | 471
en (i)
«]
Offs
pis
0]
Contralter output
Proportional reponse
Derivative responte
ol >;
FIGURE 9.21
Proportional-derivative (PD) action showing the offset error from the proportional mode.
This example is for reverse action.
Solution
In this case, we evaluate
dep
P= Krey + KoKp "+ pp
‘over the (wo spans ofthe error. In the time of Oto I s where e,
Di = Kyat + KpKpa + pp
or, because a = 1%%/s,
p= +25 +20
Note the instantaneous change of 2.5% produced by this error. In the span from 1 to 3 s,
we have
P= S420 =25 472 | CHAPTER 8
(scones)
Controtter ovtout
28
3
) Controller output
FIGURE 9.22
Solution for Example 9.9,
The span from 3 to 5 shas an error of ¢, = —0.51 + 2.5, so that we get for 3 105 s
St + 12.5 ~ 12.5 + 20
or
Py = 251 + 31.25
‘This controlled output is plotted in Figure 9.22b.
9.6.3 Three-Mode Controller (PID)
One of the most powerful but complex controller mode operations combines the propor-
tional, integral, and derivative modes. This system can be used for virtually any process
condition. The analytic expression is
de,
P= Key + KpK, | epdt + KpKy |
+ pi(0) 21) EXAMPLE
9.10
CONTROLLER PRINCIPLES | 473
ep (8)
0
Error
WW
ps
1100]
Derivative
Preprtor tag
wk
Controtier output
°
FIGURE 9.23
‘The three-mode controller action exhibits proportional, integral, end derivative action,
where all terms have been defined eaulier.
‘This mode eliminates the offset of the proportional mode and still provides fast re-
sponse. In Figure 9.23, the response of the three-mode system to an error is shown,
Let us combine everything and see how the eror of Figure 9.22a produces an output in the
three-tnode controller with K'p = 5, Ky = 0.78"!, Ky = OSs, and p,(0) = 20%. Draw
a plot of the controller output,
Solution
From Figure 9.22 the error can be expressed as follows;
Os a%
13s c= 1%
1
3-58 att 25% 474 | CHAPTER 8
‘We must epply each of these spans to the three-mode equation for controller output
P= Kpey + KK, f
dey
dt + KpKp Ge * Pi(0)
' de
or
py = St + LISP + 225
‘This is plotted in Figure 9.24 in the span of Oto Ls. At the end of 1, the integral term has
accumulated to py(1) = 21.75%. Now, from 1 t0 3s, we have
nose3sf (dt + 21.75
pz = 35(¢ — 1) + 26.79
Controlter outout (8)
1 teecond)
FIGURE 9.24
Solution for Example 8.10. CONTROLLER PRINCIPLES | 475
‘This controller variation is shown in Figure 9.24 from 1 to 3 s, At the end of 3s, the inte-
agral term has accumulated to a value of p,(3) = 28.75%. Finally, from 3 t0 5 s, we have
1428) 35 [hes 25)a—2 sa
Py = -O875" + 6.251 + 21.625
“This is plotted in Figure 9.24 from 3105 s. After 5 s, the errr is zero. Therefor, the ovt-
put will simply be the accumulated integral response providing a constant output of
Py = 32.25%.
‘The examples used in this chapter are idealized in terms of the sudden way that er-
rors change. In the real world, changes are not instantaneous, and therefore the sharp breaks
{in output, such as those shown in Figure 9.24, do not occur.
9.6.4 Special Terminology
‘A number of special terms are used in process control for discussing the controller modes.
“The following summary defines some of these terms and shows how they relate to the equa-
tions presented in this chapter.
1, Proportional band (PB) Although this term was defined earlier, let us note again
that this is the percentage error that results in 8 100% change in controller output.
2. Repeats per minute This term is another expression of the integral gain for PI
and PID controller modes. The term derives from the observation that the inte-
agral gain, Kj, has the effect of causing the controller output to change every unit,
time by the proportional mode amount. You can also see this by taking the de-
rvative of the integral term in the controller equation. This gives a change in
controller output Ap of
Ap = KiKre,At
Because Ke, is just the proportional contribution, in a unit time interval
Af = 1, K; just repeats the proportional term. For example, if = 0.5% and
Kp = 10%, then Kpey = 5%. 1f K, = 10%/(%-tin), then every minute, the
cutput would increase by 5% times 10% /(Yé-min) or 50% or 10 repeats per
minute.” It repeats the proportional amount 10 times per minute
3. Rate gain This is just another way of saying the derivative gain, Kp. Because
Kp has the units of %-5/% (or %-min/%), one often expresses the gnin as time
diccctly, Thus, arate gain of 0.05 min ora derivative time of 0.05 min both mean,
Kp = 0.05%-min/%
4. Direcu/everse action This specifies whether the controller output should in-
crease (direct) or decrease (reverse) for an increasing controlled variable, The ac-
tion is specified by the sign of the proportional gain; Ky <0 is direct, and
Kp > Oisreverse 476 | CHAPTER 9
SUMMARY
‘This chapter covers the general characteristics of controller operating modes without con-
sidering implementation of these functions. Numerous terms that are important to an un-
derstanding of controller operations are defined. The highlighted items are as follows:
1. In consideting controller operating modes, it is important to know the process load,
which is the nominal value of all process parameters, and the process lag, which rep.
resents a delay in teaction of the controller variable to a change of load variable.
2. Some processes exhibit self-regularion—that is, the characteristic that a dynamic vari-
able adopts some nominal value commensurate with the load with no control action,
3. The controller operation is defined through a relationship between percentage error or
deviation relative to full scale
= x 100 (03)
6,
and the controller output as a percentage of the controlling parameter
a= x 100 oa)
4. Control lag and dead time, respectively, refer to a delay in controller response when a
