Control: Rolta Academy:Engineering Design Services

Rolta Academy :Engineering Design Services
CHAPTER 6
CONTROL
6.0 INTRODUCTION
Control of the processes in the plant is an essential part of the plant operation.
There must be enough water in the boilers to act as a heat sink for the reactor but
there must not be water flowing out the top of the boilers towards the turbine. The
level of the boiler must be kept within a certain range. The heat transport
pressure is another critical parameter that must be controlled. If it is too high the
system will burst, if it is too low the water will boil. Either condition impairs the
ability of the heat transport system to cool the fuel.
In this section we will look at the very basics of control. We will examine the
fundamental control building blocks of proportional, integral and differential and
their application to some simple systems.
6.1BASIC CONTROL PRINCIPLES
Consider a typical process control system. or a particular example let us look at
an open tank, which supplies a process, say, a pump, at its output. The tank will
re!uire a supply to maintain its level "and therefore the pump#s positive suction
head$ at a fixed predetermined point. This predetermined level is referred to as
the set point "%&$ and it is also the controlled !uantity of the system.
Clearly whilst the inflow and outflow are in mass balance, the level will remain
constant. 'ny difference in the relative flows will cause the level to vary. (ow can
we effectively control this system to a constant level) We must first identify our
variables. *bviously there could be a number of variables in any system, the two
in which we are most interested are+
The controlled variable , in our example this will be level.
The manipulated variable - the inflow or outflow from the system.
If we look more closely at our sample system "ig ../$, assuming the level is at
the set point, the inflow to the system and outflow are balanced. *bviously no
control action is re!uired whilst this status !uo exists. Control action is only
necessary when a difference or error exists between the set point and the
measured level. 0epending on whether this error is a positive or negative !uantity,
the appropriate control correction will be made in an attempt to restore the
process to the set point.
(enceforth, the error will always take the form of+
Error 1 %et point - 2easured 3uantity
e 1 %& , 2
Instrumentation Engineering
*4
1
"../$
Fig 6.1 Level Control Syte!
The control action will be either to vary the inflow or outflow from the system in
order to keep the level at the set point. 5et us consider the general format for
achieving these ob6ectives.
's can be seen from ig ..7, the process can be represented by a closed loop.
The system output "level$ is monitored by a process sensor and the measurement
signal is feedback to a comparator at the input of the system. The second input
to the comparator is the set point signal8 the comparator#s output being the
difference or error signal. The amplifier, a present 6ust a black box, will provide the
appropriate correction to maintain the process at its set point despite
disturbances that may occur. It can be seen that if the system were being
operated in manual control the feedback path would not be present. The operator
would provide this feedback and apply the necessary correction to the system
whilst observing the effect on the controlled variable. This is termed open loop
operation.
Fig 6."#
CLOSED LOOP PROCESS
6.1.1 Fee$%&'( Control
This concept 6ustifies the use of the word negative in three ways+
2
The negative aspect of feeding the measured signal backwards from the
output to the input of the system. "'ctual definition of negative feedback
control$.
The control correction must be negative in that a correction rather than a
compounding of error must occur.
The fact that an error must occur before a correction can take place, i.e.,
retrospective or negative control action.
In the next section we will study in more detail the methods used to effect the
necessary control corrections.
6.1." Fee$ )or*&r$ Control
If we wish to control our process without an error first occurring, we must base
our control on correction of the disturbances, which will eventually, cause a
process error. This is termed feed forward control. eed forward control is rarely
if ever used on its own but is used in con6unction with feedback control to improve
the response of control to process disturbances.
6.1.+ S,!!&ry
Controlled 9ariable - output !uantity of system "5evel, Temperature, etc.$.
2anipulated 9ariable - means of maintaining controlled variable at the set
point.
Error signal - e!uals the difference between the set point and the
measurement. "e 1 %& - 2$.
%et point - desired process level. "%&$
2easurement - actual process level. "2$
Closed 5oop - automatic control.
*pen 5oop - manual control.
Feedback control is error correction following a disturbance.
Feed forward control is control of disturbances, which could cause a process error.

6."ON-OFF CONTROL
3
Fig 6 + Ty.i'&l On-O)) Control S'/e!e
5et us consider our level control system in a little more practical detail. The valve
in the inflow line to the system is an electrically operated solenoid valve.
"4emember an electrically operated solenoid valve has only two operating
positions - fully open or fully closed.$ 'ssume that under initial conditions with a
demand on the system the level will start to fall and 9
/
will have to be opened to
provide an inflow. This can easily be achieved by mounting a differential pressure
switch, &/ at the bottom of the tank to operate when the level falls to 5
/
. When the
level is at 5
/
the li!uid will be height h
/
above switch. The pressure at the switch
will be &
/
1 :gh
/
.
Where : - the mass density of the li!uid
g - the acceleration due to gravity
h
/
- the height of the li!uid
The resulting switch closure can energi;e the solenoid valve 9
/
causing an inflow
to the tank. 'ssuming the valve is correctly si;ed, this will cause a rise in the
level back towards the set point.
In order to arrest the rise in level the built in differential feature of the switch can
be employed to de,energi;e the solenoid valve when level 5
7
is reached. This
system will achieve a mean level in the tank about the desired set point. This
method is known as *<=* control. Clearly it is impossible to maintain the
system at the set point since there must be a difference in the operating levels 5
/
and 5
7
as the valve can only be energi;ed or de,energi;ed. It is often counter
productive to try to reduce the differential between 5
/
and 5
7
to too small a value
4
as this will result in excessive cycling, and hence wear, of the valve. >sual
practice is to control with a dead band about the set point as shown in ig ..?.
Fig 6.0 Ty.i'&l ON-OFF Re.one
The sinusoidal cycling is typical of on=off control. on=off control can be used to
advantage on a sluggish system, i.e., where the periodic time is large. Typical
uses in C'<0> units are electric heater controls in de,aerator tanks and &(T%
bleed condenser and pressuri;er. If fine control is re!uired a simple on=off control
system is inade!uate. We will discuss a method for achieving a finer control in
the next section.
6.".1 S,!!&ry
@ *n=off control , control signal is either AB or /AAB
@ Control at set point not achievable, a dead band must be incorporated.
@ >seful for large, sluggish systems particularly those incorporating electric
heaters.
6.+ BASIC PROPORTIONAL CONTROL
In our example of on=off control it was seen that an all or nothing control correction
was applied as the result of an error signal occurring. Clearly it would be to our
advantage if the control signal were proportional to the magnitude of error. This is
the basis of proportional control and is the most fre!uently encountered control
mode. (ow can this control be achieved) 4eferring to ig ..C it can be seen that we
can modify our system to use a pneumatically operated control valve and a level
transmitter with a 7A - /AA k&a pneumatic output.
5
Fig 6.1 Level Control o) O.en T&n(
If the outflow "3
o
$ increases then the level in the tank will fall. The pressure sensed
by the level transmitter, which is representative of the level in the tank, will also fall
causing a decrease in the output signal from the level transmitter. This output signal
is fed to the "air to close$ control valve "valve fully open with 7A k&a signal, fully
closed with /AA k&a signal$. ' falling level will therefore cause the valve to
progressively open and hence raise the level in the tank. The system as shown is
somewhat impractical as the initial set point conditions will need to be set by some
manual method and then ensuring that steady state conditions are achieved with
the valve at, say CAB opening and a level transmitter output of .A k&a "CAB range$.
This simple system does illustrate however a ma6or disadvantage with proportional
control. <otice that the control signal "valve opening$ can only change when the
level signal is changing. Thus if a disturbance occurs, say an increase in demand,
the level will drop and the output from the level transmitter will also fall. This will
cause the air to close valve to open more, hence increasing the inflow.
'fter a period of time the inflow will have increased such that a now mass balance is
established between inflow and outflow. Dut where is the level at this time)
Certainly not at the set point. In the example given it will stabili;e at some steady
state level below the set point. This steady state deviation is known as offset and is
inherent in all proportional control systems. 0espite this obvious disadvantage, "we
cannot return the process to the set point after a disturbance with proportional
control$ this mode of control will form the basis for all our control strategies. In the
next section we will discuss a more practical control scheme using proportional
control and also ways of lessening the problem of offset.
Example /
' tank has inflow and outflow e!ual to CAB of maximum and its level is at the set
point, say CAB. ' step change in outflow occurs to .AB "E/AB$. *utflow now
exceeds inflow so the level will fall. The output from the level transmitter will also fall
and, for our system, will match the fall in level - say /B change in signal for a /B
change in level. The 5T signal will open the '=C valve more, by /B in fact. The
inflow is now C/B, still less than the outflow. The level will continue to fall until inflow
e!uals outflow, i.e., ".AB$. This can only happen when the 5T signal has changed
by /AB$ and this change reflects a drop in level on /AB+ i.e., /AB offset.
To restore the process to the set point re!uires a further increase of inflow. This
increase can only be achieved by a further decrease in signal to the valve "i.e., as
decrease in 5T output corresponding to a further decrease in level$.
With the conditions as stated in the example there is no way in which a CAB level
can be achieved with a .AB outflow. ' CAB level with a .AB outflow re!uires a .AB
inflow. *ur systems can only provide a .AB inflow from a ?AB level signal.
6
Example 7
'n alternative method of illustrating proportional control is by means of a simple float
system "ig ...$. 'ssume the inflow and outflow are e!ual and the level is at the set
point. If an increase in outflow occurs the level in the tank must fall. The float will also
fall as the level falls. This drop in float position will cause the valve on the inflow to
open more thus increasing the inflow. Eventually the fall in level will result in a valve
opening, which will restore the mass balance between the in flow and the outflow
<ote an increased inflow can only be achieved as a result of a lower level in the
tank. The level is no longer at the set point an offset has been generated
Fig 6.6 Si!.le Pro.ortion&l Controller
6.+.1 S,!!&ry
&roportional control provides a control signal, proportional to the magnitude
and direction of the error signal.
'fter a disturbance, proportional control will provide only a new mass balance
situation. ' change in control signal re!uires a change in error signal,
therefore offset will occur.
&roportional control stabili;es an error8 it does not remove it.
6.0Pro.ortion&l Control
6.0.1 Ter!inology
7
6.0." Pr&'ti'&l Pro.ortion&l Control
' more practical proportional control scheme can be achieved by inserting a
controller between the level transmitter and the control valve. This will eliminate
the setting up problems mentioned in the previous module "i.e., it will have a set
point control$ and also introduce other advantages, which will be discussed in
this section.
In a practical system one of the primary considerations is the failure mode of the
valve.
In our example of an open tank with a valve on the inflow it would be reasonable
to assume that the valve should close in the event of an air supply failure to
prevent the tank overflowing, i.e., an air to open valve
8
Fig 6.2 O.en T&n( Control
To achieve the necessary control action on, say, a falling tank level it is necessary
to convert the decreasing output of the level transmitter to an increasing input
signal to the control valve. The level controller will perform this function and is
termed an indirect or reverse acting "FG$ controller. It can be seen that if the valve
action had been chosen air to close, then this reversal would not have been
re!uired and a direct "FF$ acting controller could have been used. <ormally
controllers are capable of performing either control action, direct or reverse, by a
simple switching process.
The controller will also accept our desired set point input and perform the
comparison between set point and measurement to calculate the error#s
magnitude and direction.
>p to now we have only assumed proportionality constant or one, i.e., the control
signal e!uals the input error. Is this always the best ratio) Consider the following
graphs of input, output and level with respect to time+
Fig 6.3 Pro.ortion&l Control Re.one C,rve

Fig 6.4 Pro.ortion&l Re.one *it/ lo*er Pro.ortion&l B&n$
It can be seen that a step increase in demand "outflow$ has occurred at time t
A
.
the resulting control correction has caused a new mass balance to be achieved
after some time t/. 't this time, under the new mass balance conditions, the level
will stabili;e at some level below the original set point, i.e., an offset has
occurred, the loss in volume being represented by the shaded area between the
input and output curves.
Consider now the same demand disturbance but with the control signal increased
in relative magnitude with respect to the error signal8 i.e., instead of control signal
1 error signal, control signal 1 error signal x gain constant "k$. Clearly for any
given error signal the control signal will be increased in magnitude, the inflow will
be increased, and a new mass balance will be achieved in a shorter time as
shown in ig ..H. "If we refer back to our simple ball cock system in section I.I, it
can be seen that the gain could be varied by ad6usting the position of the valve,
operating link on the float arm.$ The offset is much reduced. In instrumentation
this ad6ustment of controller gain is referred to as proportional band "&D$.
&roportional band is defined as that input signal span change, in percent, which
will cause a hundred percent change in output signal.
or example if an input signal span change of /AAB is re!uired to give an output
change of /AAB the system is said to have a proportional band of /AAB. If the
system was now ad6usted such that the /AAB change in output was achieved
with only a CAB change in input signal span then the proportional band is now
said to be CAB. There is a clear relationship between proportional band and gain.
Jain can be defined as the ratio between change in output and change in input.
1!
Dy inspection it can be seen that a &D of /AAB is the same as a gain of one since
change of input e!uals change in output. &D is the reciprocal of gain, expressed
as a percentage. The general relationship is+

%mall values of &D "high gain$ are usually referred to as narrow proportional band
whilst low gain is termed wide proportional band. <ote there is no magic ig ../A to
define narrow or wide proportional band, relative values only are applicable, for
example, /CB &D is wider than /AB &D, /CAB &D is narrower than 7AAB &D.
Fig 6.10 Re.one 5er, PB6 Pro.ortion&l Control Only
We have seen from the two earlier examples that increasing the gain "narrowing the
&D$ caused the offset to be decreased. Can this procedure be used to reduce the
offset to ;ero)
11
Fig 6.11 7 De'&y Re.one C,rve
With reference to ig ../A, consider a high gain system "say gain 1 CA, &D 1 7B$.
>nder steady state conditions with the process at the set point the inflow will have a
constant value. This is usually taken to be a control signal of CAB for a proportional
controller with the process at the set point. In other words we have a CAB control
capability. With our high gain system it can be seen that the maximum control signal
will be achieved with an error of 1/B "control signal 1 gain x error$. This control
signal will cause the valve to go fully open, the level will rise and the process will
cross the set point. The error signal will now change sign and when the error again
exceeds /B the resultant control signal will now cause the valve to fully close hence
completely stopping the inflow. This process will be repeated continuously , we have
reverted to an on=off control situation with all the disadvantages previously
mentioned. *bviously there must be some optimum setting of &D which is a trade off
between the highly stable but sluggish low gain system with large offset, and the
fast acting, unstable on=off system with mean offset e!ual to ;ero. The accepted
optimum setting is one that causes the process to decay in a K decay method as
shown in both ig ../A and ..//.
The !uarter decay curves show that the process returns to a steady state condition
after three cycles of damped oscillation. This optimi;ation will be discussed more
fully in the section on controller tuning.
4ecall the output of a proportional controller is e!ual to+

m 1 ke where m 1 control signal
k 1 controller gain 1 /AAB=&D
e 1 error signal 1 "%& , 2$
Clearly if the error is ;ero the control signal will be ;ero, this is an undesirable
situation. Therefore for proportional control a constant term or bias must be added to
provide a steady state control signal when the error is ;ero.
or the purposes of this course we will assume the steady state output of a
proportional controller when at the set point to be CAB. The e!uation for
proportional control becomes+
m 1 ke E b where b 1 bias "1CAB added to output signal$
C&l',l&tion o) o))et
E8&!.le#
'n air to open valve on the inflow controls level in the tank. When the process is at
the set point the valve opening is CAB .'n increase in outflow results in the valve
opening, increasing to a new steady state value of LAB. What is the resulting offset
if the controller &D is +
a$ CAB
b$ 7CB
12
'nswer+ to achieve correct control the controller will be reverse "FG$ acting.
a$ &D1CAB gain 17
Change in valve position 1LA,CA17AB
This is the output change for the controller.
Jain1output=input
717AB=input
Input1/AB
%ince the controller is reverse acting measured variable must have been negative
i.e.8 -/AB. This is e!ual to a E error or a - offset
*ffset1 ,/AB below set point.
b$ &D17CB gain 1?
Input 1CB
*ffset 1 ,CB below set point
<ote that the narrower &D is likely to introduce some degree of oscillation in to the
system. (opefully this will be a damped oscillation.
6.0.+ S,!!&ry
The controller action must be chosen "either direct. or reverse "FG$$ to
achieve the correct control response.
&roportional Dand 1 /AAB=gain or gain 1 /AAB=&D

The optimum settings for &D should result in the process decaying in a K
decay mode.
6.1 Reet o) Integr&l A'tion
2ost of the processes we will be controlling will have a clearly defined set point. If
we wish to restore the process to the set point after a disturbance then proportional
action alone will be insufficient.
Consider again the diagram "ig ../7$ showing the response of a system under
proportional control.
13
Fig 6.1" Re.one C,rve6 Pro.ortion&l Control Only
If we wish to restore the process to the set point we must increase the inflow over
and above that re!uired restoring a mass balance. The additional inflow must
replace the lost volume and then revert to a mass balance situation to maintain the
level at the set point. This is shown in ig ../I. This additional control signal must
be present until the error signal is once again ;ero.
Fig 6.1+ A$$ition&l Control Sign&l Retore Pro'e To Set .oint
This additional control signal is known as 4eset action, it resets the process to the
set point. 4eset action is always used in con6unction with proportional action.
2athematically, reset action is the integration of the error signal to ;ero hence the
alternative nomenclature - Integral action.
The combination of proportional plus reset action is usually referred to as &I control.
The response of &I control is best considered in open loop form, i.e., the loop is
opened 6ust before the final control element so that the control correction is not in
fact made. This is illustrated in ig ../?.
Fig.6.10 Pro.ortion&l Pl, Reet6 O.en Loo. Re.one
14
It can be seen that proportional action will be e!ual to ke where k is the gain of the
controller. 4eset action will cause a ramping of the output signal to provide the
necessary extra control action. 'fter time, say t, the reset action has repeated the
original proportional response8 this is the repeat time, the unit chosen for defining
reset action. It can be seen that increased reset action would increase the slope of
the reset ramp.
<ote that proportional action occurs first followed by reset action.
4eset action is defined as either reset rate in repeats per minute "4&2$ or reset time
in minutes per repeat "2&4$.
We have already mentioned that the optimum setting for proportional control is one,
which produces a K decay curve. What is the optimum setting for reset action) We
will discuss this more fully in the module on controller tuning. or now, let us 6ust
consider a very slow reset rate and a very fast reset rate.
15
' very slow reset rate will ramp the control signal up very slowly. Eventually the
process will be returned to the set point. The control will be very sluggish and if the
system is sub6ected to fre!uent disturbances the process may not ever be fully
restored to the set pointM
If a very fast reset rate is used, the control signal will increase very !uickly. If we are
controlling, say, a large volume tank, the level response of the tank may lag behind
the response of the controller.
The control signal will go to its limiting value "A or /AAB$ and the limiting control
signal will eventually cause the process to cross the set point. The error signal will
now change its sign, and reset action will also reverse direction and !uickly ramp to
the other extreme.
This process will continue indefinitely, the control valve cycling, with resulting wear
and tear, from one extreme to the other. The actual process level will cycle about
the set point. This cycling is known as reset windup and will occur if the process is
sub6ect to a sustained error and a too fast reset rate. The reset rate must be
decreased "reset time increased$.
The mathematical expression for & E I control becomes+
&roportional control i.e., "proper sign of gain$ inputs a /NAO lag into the system "the
correction must be opposite to the error$. 4eset action introduces a further lag. This
fact must be taken into account when tuning the controller. "It follows proportional
action$. The total lag must be increased and is now closer to I.AO. "I.AO lag means
the feedback signal is now in phase with the input and adding to it , the system is
now unstable.$ 4eset action causes the loop to be less stable.
6.1.1 S,!!&ry
4eset action removes offset.
It#s units are 4epeats per 2inute "4&2$ or 2inutes per 4epeat "2&4$
16
If reset action is faster than the process can respond, 4eset Windup can
occur.
4eset 'ction makes a control loop less stable
0o not sub6ect process loops with reset control to sustained errors - the
control signal will be ramped to the extreme value , reset windup will occur.
6.6 RATE OR DERI5ATI5E ACTION
Consider a control system sub6ected to a disturbance, which causes the error to
increase in a ramped manner. &roportional control would respond to this ramped
error with a similarly ramped output signal whose slope is proportional to the
controller gain. We could reduce the final deviation from the setpoint, i.e., the offset,
and the recovery time, if we can provide some extra control signal related to the rate
of change of the error signal. This is termed rate or derivative action and is usually
incorporated with proportional control.
4ate action is an anticipatory control, which provides a large initial control signal to
limit the final deviation. The typical open loop response is shown in ig . /C.
It can be seen that the derivative action gives a large, immediate, control signal,
which will limit the deviation. &roportional action is then superimposed upon this
step. When the error stops changing derivative action ceases. <ote that the
displayed step response unobtainable in practice because the normal response
approximates and exponential rise and decay.
The rate response gives an immediate control signal, which will be e!ual to what the
proportional response would be after some time, say, T minutes. 0erivative units are
given in minutes. These are the minutes advance of proportional action. 0erivative
action is a leading control and, therefore, tends to reduce the overall lag in the
system - the system is somewhat more stable.
Fig 6 11 Pro.ortion&l &n$ Deriv&tive9O.en Loo. Re.one
17
The use of derivative control is limited. 't first glance, derivative control looks
attractive. It should help reduce the time re!uired to stabili;e an error. (owever, it
will not remove offset. The control signal from derivative action ceases when the
error stops changing, which will not necessarily be at the set point.
Its use, in practice, is also limited to slow acting processes. If used on a fast acting
process, such as flow, control signals due to derivative action will often drive the
control valve to extremes following !uite small but steep "large de/dt$ changes in
input.
Consider a simple flow control system, consisting of an orifice plate with flow
transmitter and s!uare root extractor plus direct acting controller and air to close
valve "refer to ig . /.$. This system is sub6ected to a small, but fast, process
disturbance. (ow will this control scheme perform under proportional and derivative
control modes)
Fig 6.16 Si!.le Flo* Control Syte!
To answer this !uestion let us consider the &0 response a fast change in process
%ignal in an open loop system "ig . /L$.
18

Fig 6.12 O.en Loo. Re.one O) Pro.ortion&l Pl, Deriv&tive :PD; A'tion to
R&.i$ly C/&nging Error Sign&l
The upper portion of ig ./L shows a positive process excursion, 'D, from the ;ero
error condition, followed by an e!ual negative excursion, DC, which returns the error
to ;ero. <ote that the rate of change, i.e., the slope of the process change, from D to
C is twice the rate of change of the process, from ' to D. 2athematically+
The proportional control action from D to C will be e!ual but opposite to the
proportional control action from ' to D. The rate or derivative control action from D to
C will be double that from ' to D. The resulting open loop control signal pattern is
shown in the lower portion of ig ../L. The controller gain and derivative settings
remain constant.
9ery shortly after time "t
A
$ the control signal increases abruptly to a value determined
by the rate of change of the error "e$, the derivative or rate time setting, and the
controller gain. &roportional action ramps the control signal up, until time "t
/
$, to a
value determined by the error "e$ and the controller gain setting. This includes the
direction of the error and controller action.
't time "t
/
$ the rate of change of the process error, de/dt, momentarily becomes ;ero,
so the original change in the control signal due to the rate action drops out. Then,
the process error change direction becomes negative, and the derivative control
action now produces an abrupt negative control signal, double the original derivative
control signal. The proportional control action then ramps the control signal down
until time "t
7
$
1
"..7$
't time "t
7
$ the rate of change of the process error becomes ;ero, so the derivative
control signal again drops out leaving the control signal at its original bias ";ero$
error value. <ote that this final bias, ";ero$ error value of the control signal and,
(ence, the control valve position at the end of this excursion is determined solely by
the proportional. The valve has been stroked rapidly and repeated by the derivative
action sub6ecting it to unnecessary wear, with no improvement in control.
The response of the closed loop shown in ig ../. would be somewhat different
because the resulting valve action would continuously alter the error signal.
(owever, the valve would still be sub6ected to rapid and repeated stroking
unnecessarily.
Thus, it can be seen from the above discussion that the use of derivative action on
fast acting processes such as flow is not advisable.
5et us look at the control of a sluggish "generally a physically large$ system. 's an
example, consider a large tank with a variable outflow and a control valve on the
inflow. ' large volume change will, therefore, be necessary before any appreciable
change in level occurs.
Consider a large change in the outflow. 'fter some delay "due to the sluggishness of
the system$ the controller will respond.
If we have only proportional mode on the controller the delays will mean that the
controller is always chasing the error initiated by the outflow disturbance. The
response to proportional control is shown in ig ../N <ote that the process has not
fully stabili;ed after a considerable period of time.
The addition of derivative action, however, causes an anticipatory response. The
control signal increases more rapidly and the process is returned to a steady state in
a much shorter time. <ote also that+
The system is more stable "less cycling$ with &0 control. *ffset still exists.
Fig 6.13 L&rge Syte! Un$er .ro.ortion&l &n$ Pro.ortion&l Pl, Deriv&tive
Control
6.2 <ULTIPLE CONTROL <ODES
We have already discussed some of the possible combinations of control modes.
These are+
&roportional only,
&roportional plus reset "integral$ & E I,
2!
&roportional plus derivative "rate$ & E 0.
It is also possible to use a combination of all three,control modes,
&roportional plus Integral plus 0erivative "& E I E 0$.
't a glance proportional only does not appear very attractive - we will get an offset
as the result of a disturbance and invariably we wish to control to a fixed set point.
'n application of proportional only control in a C'<0> system is in the li!uid ;one
level control system. The reason that straight proportional control can be used here
is that the controlled variable is not level but neutron flux. The manipulated variable
is the water level8 therefore offset is not important as the level is manipulated to
provide the re!uired neutron flux.
In general it can be said that the vast ma6ority of control systems "probably greater
than HAB$ will incorporate proportional plus integral modes. "We usually want to
control to a fixed set point.$ low control systems will invariably have & E I control.
0erivative control will generally be limited to large sluggish systems with long
inherent control time delays, "for example, that shown in ig . /N.$. ' good general
example is the heat exchanger. The thermal interchange process is often slow and
the temperature sensor is usually installed in a thermal well, which further slows the
control signal response. re!uently heat exchanger temperature controllers will
incorporate three,mode control "& E I E 0$

6.3T=PICAL NE>ATI5E FEEDBAC? CONTROL SCHE<ES
6.3.1 Level Control
In general we can divide level measurement into three types+ *pen Tanks Closed
Tanks
B,%%ler Syte! :O.en or Cloe$ T&n(;
If a differential pressure transmitter is used as a level detector, the low,pressure port
will be vented to atmosphere in an open tank application. In a closed tank, where
there is often a gas phase at pressure above the li!uid, the low,pressure port will be
taken to the top of the tank. 'ny gas pressure will then be e!ually sensed by the
high and low sides and thus cancelled. 4emember the closed tank installation will
have either a wet or dry leg on the low,pressure sides.
O.en T&n( Int&ll&tion
'ssuming the control valve is on the inflow, the best failure mode for the valve
would be to fail closed, i.e., 'ir to *pen "'=*$ valve. The pressure sensed at the
base of the tank on a falling level will decrease, i.e., controller input. The valve must
open more, to replenish the tank, re!uiring an increasing signal. The controller must
21
be reverse acting and will usually have & E I modes. The system is shown in ig .
/H
If it is necessary to mount the valve in the outflow, the best failure mode would
probably be to fail open "'=C$. This valve action would re!uire an increasing signal to
halt a falling tank level8 again a reverse acting "& E I$ controller is necessary.
The same reasoning would apply to closed tank or bubbler systems, the only
difference being in the sensing method employed. 4emember control modes use of
derivative action on large, slow, systems.
Fig 6.14 O.en T&n( Level Control
Flo* Control
Fig 6."0 Ty.i'&l Flo* Control
' typical flow control system re!uires some form of restriction to provide a pressure
differential proportional to flow "e.g. orifice plate$ plus a s!uare root extractor to
provide a linear signal. The controller action depends upon the choice of control
valve. If an air to open valve is chosen then controller action should be reverse, as
an increase in flow must be countered by a decrease in valve opening. or an air to
close valve the action must of course be direct. The general format is shown in ig
. 7A.
The control modes will be proportional plus integral "never use derivative on a
flow control loop$.
22
6.3.+ Pre,re Control
The control of pressure in, say, a pressure vessel, is generally achieved in one of
three ways.
1. 9ariable eed with Constant Dleed
2. Constant eed with 9ariable Dleed
3. 9ariable eed and Dleed
Consider first 9ariable eed and Constant Dleed "ig . 7/$. The feed valve
action is air to close "'=C$. Increasing pressure will re!uire an increasing valve
signal to throttle the supply. The "& E I$ controller is direct acting. or a variable
bleed application the control valve will be transferred to the bleed application the
control valve will be transferred to the bleed line and will need to be '=* if a
direct acting controller is used.
Fig 6."1 Pre,re Control Cont&nt Blee$
or variable feed and bleed we can use a split range control scheme "one
controller driving two valves$. This is shown in ig . 77. When at the set point we
re!uire feed to e!ual bleed. If pressure increases we re!uire less feed action and
more bleed action and vice versa. The valve actions must therefore be opposite,
say feed valve '=C and bleed valve '=*. *n increasing pressure the direct acting
controller will supply a larger signal to the feed valve "closing it$ and
to the bleed valve "opening it$. &ressure should thus be maintained at the set
point with proportional plus integral control.
Fig 6."" S.lit R&nge )ee$ &n$ Blee$ Pre,re Control
6.3.0Te!.er&t,re Control
23
The general problem with temperature control is the slowness of response. or
this reason the use of derivative action is fairly standard. ig ..7I shows a
representative heat exchanger, which cools hot bleed with cold service water.
The choice of control valve would probably be air to close, i.e., fail open, to give
maximum cooling in the event of a air supply failure to the valve.
Fig 6."+ Te!.er&t,re Control o) He&t E8'/&nger
'n increase, say, in bleed temperature re!uires a larger valve opening, i.e.,
smaller valve signal. ' reverse acting controller is re!uired. Three mode, & E I E
0, control is fairly usual.
6.4 <ULTI5ARIABLE-AD5ANCED CONTROL LOOPS
Multivariable loops are control loops in which a primary controller controls one
process variable by sending signals to a controller of a different loop that impacts
the process variable of the primary loop. or example, the primary process
variable may be the temperature of the fluid in a tank that is heated by a steam
6acket "a pressuri;ed steam chamber surrounding the tank$. To control the
primary variable "temperature$, the primary "master$ controller signals the
secondary "slave$ controller that is controlling steam pressure. The primary
controller will manipulate the set point of the secondary controller to maintain the
set point temperature of the primary process variable "ig ..7?$.
Fig 6."0
<,ltiv&ri&%le Loo.
24
When tuning a control loop, it is important to take into account the presence of
multivariable loops. The standard procedure is to tune the secondary loop before
tuning the primary loop because ad6ustments to the secondary loop impact the
primary loop. Tuning the primary loop will not impact the secondary loop tuning.
6.4.1 FEEDFOR@ARD CONTROL
Feed forward control is a control system that anticipates load disturbances and
controls them before they can impact the process variable. or feed forward
control to work, the user must have a mathematical understanding of how the
manipulated variables will impact the process variable. ig ..7C shows a feed
forward loop in which a flow transmitter opens or closes a hot steam valve based
on how much cold fluid passes through the flow sensor.
Fig 6."1 Fee$ For*&r$ Control
'n advantage of feed forward control is that error is prevented, rather than
corrected. (owever, it is difficult to account for all possible load disturbances in a
system through feed forward control. actors such as outside temperature,
buildup in pipes, consistency of raw materials, humidity, and moisture content
can all become load disturbances and cannot always be effectively accounted for
in a feed forward system.
In general, feed forward systems should be used in cases where the controlled
variable has the potential of being a ma6or load disturbance on the process
variable ultimately being controlled. The added complexity and expense of feed
forward control may not be e!ual to the benefits of increased control in the case
of a variable that causes only a small load disturbance.
6.4." FEEDFOR@ARD PLUS FEEDBAC?
Decause of the difficulty of accounting for every possible load disturbance in a
feed forward system, feed forward systems are often combined with feedback
systems. Controllers with summing functions are used in these combined
systems to total the input from both the feed forward loop and the feedback loop,
and send a unified signal to the final control element. ig ..7. shows a feed
25
forward,plus,feedback loop in which both a flow transmitter and a temperature
transmitter provide information for controlling a hot steam valve.
Fig 6."6 Fee$ )or*&r$ .l, Fee$%&'( Control Syte!
This module has discussed specific types of control loops, what components are
used in them, and some of the applications "e.g., flow, pressure, temperature$
they are applied to. In practice, however, many independent and interconnected
loops are combined to control the workings of a typical plant. This section will
ac!uaint you with some of the methods of control currently being used in process
industries.
6.4.+ CASCADE CONTROL
Cascade control is a control system in which a secondary "slave$ control loop is
set up to control a variable that is a ma6or source of load disturbance for another
primary "master$ control loop. The controller of the primary loop determines the
set point of the summing controller in the secondary loop "ig ..7L$.
26
Fig 6."2 C&'&$e Control
6.4.0 BATCH CONTROL
Batch processes are those processes that are taken from start to finish in
batches. or example, mixing the ingredients for a 6uice drinks is often a batch
process. Typically, a limited amount of one flavor "e.g., orange drink or apple
drink$ is mixed at a time. or these reasons, it is not practical to have a
continuous process running. Datch processes often involve getting the correct
proportion of ingredients into the batch. 5evel, flow, pressure, temperature, and
often,mass measurements are used at various stages of batch processes.
' disadvantage of batch control is that the process must be fre!uently restarted.
%tart,up presents control problems because, typically, all measurements in the
system are below set point at start,up. 'nother disadvantage is that as recipes
change, control instruments may need to be recalibrated.
6.4.1 RATIO CONTROL
Imagine a process in which an acid must be diluted with water in the proportion
two parts water to one part acid. If a tank has an acid supply on one side of a
mixing vessel and a water supply on the other, a control system could be
developed to control the ratio of acid to water, even though the water supply itself
may not be controlled. This type of control system is called ratio control (ig
..7N$. 4atio control is used in many applications and involves a controller that
receives input from a flow measurement device on the unregulated "wild$ flow.
The controller performs a ratio calculation and signals the appropriate set point to
another controller that sets the flow of the second fluid so that the proper
proportion of the second fluid can be added.
4atio control might be used where a continuous process is going on and an
additive is being put into the flow "e.g., chlorination of water$.
Fig 6."3 R&tio Control
6.4.6 SELECTI5E CONTROL
Selective control refers to a control system in which the more important of two
variables will be maintained. or example, in a boiler control system, if fuel flow
outpaces airflow, then uncombusted fuel can build up in the boiler and cause an
explosion. %elective control is used to allow for an air,rich mixture, but never a
27
fuel,rich mixture. %elective control is most often used when e!uipment must be
protected or safety maintained, even at the cost of not maintaining an optimal
process variable set point.
6.4.2 FUAA= CONTROL
Fuzzy control is a form of adaptive control in which the controller uses fu;;y logic
to make decisions about ad6usting the process. Fuzzy logic is a form of computer
logic where whether something is or is not included in a set is based on a grading
scale in which multiple factors are accounted for and rated by the computer. The
essential idea of fu;;y control is to create a kind of artificial intelligence that will
account for numerous variables, formulate a theory of how to make
improvements, ad6ust the process, and learn from the result. u;;y control is a
relatively new technology. Decause a machine makes process control changes
without consulting humans, fu;;y control removes from operators some of the
ability, but none of the responsibility, to control a process.
6.10 APPLICATION
Ai! &n$ o%Be'tive
The aim of these notes is to provide some basic ideas and rules that may be
used to select a distillation control strategy. %eparate notes will discuss more
complex mathematical techni!ues that may also be used as part of a PtoolboxP of
methods that have evolved as aids in distillation control strategy selection.
Intro$,'tion
The effective operation of a binary distillation column is determined by the control
of many variables. Jenerally, the variables in table / need to be controlled.
T&%le 6.1 Ty.i'&l 5&ri&%le t/&t /&ve to %e !&int&ine$ in & $itill&tion
Col,!n
The two main disturbances that affect a column are+
eed flow rate,
eed composition, ;
f
28
%o called Pmanipulated variablesP are ad6usted to counter,act the effect of
disturbances and ensure desired operation. Dut what are the manipulated
variables)
4elationships between inputs "mvPs and dvPs$ and outputs "cvPs$ are !uantified by
steady,state material and energy balances. To simplify preliminary discussions
consider Qperfect control# of pressure "i.e. the energy balance e!uations are not
considered$.
Ste&$yCt&te !&teri&l %&l&n'e &ro,n$ & $itill&tion 'ol,!n
The following ig ..7H is a material balance diagram for a typical distillation
column+
Fig 6."4 Ste&$y St&te <&teri&l B&l&n'e
The column feed is "kmol=min$ and the concentration of the more volatile
component in li!uid is ;
f
. The distillate flow is 0 "kmol =min$ with overhead
product concentration x
0
and the bottom product flow rate is D "kmol = min$ of
concentration x
D
.
2
or a binary column, the two independent overall balances are+
The total material balance+ 1 0 E D
The component balance+ ;
f
1 0x
0
E Dx
D
Eliminating either D or 0 from these e!uations gives the following+
The e!uations "I$ define the RcutS, i.e what percentage of the total feedflow exits
the column as distillate and bottoms product for specified inlet and outlet
concentrations.
rom e!uations "I$ it is apparent that distillate "0$ and bottoms flow "D$ are
related to top and bottom product compositions "x
0
and x
D
$ and are therefore
potential manipulated variables. 's expected, changes in and ;
f
will also affect
x
0
and x
D
.
'round the condenser and accumulator assuming a total condenser, the material
balances are#
The material balance+ 9
n
1 0 E 5
n
"..7$
The component balance+ 9
n
y
n
1 0x
0
E 5
n
x
0
"..I$
'nd around the reboiler,
The material balance+ 5
m
1 9
b
E D "..?$
The component balance+ 5
m
x
m
1 9
b
y
b
E Dx
D
"..C$
or a li!uid feed 5
m
1 5
n
E
'ssuming that the molar flows of li!uid and vapour are constant through the
column "constant molal overflow$ then,
5 1 5
n
1 5
nE/
Tetc.
9 1 9
n
1 9
nE/
1. Tetc.
Therefore+ 0 1 9 , 5 and D 1 E 5 , 9 "...$
E!uations "I$ demonstrate that 0 and D may be used to regulated x
0
and x
D
,
based upon the relationships "e!uation N$ it is obvious that 5 and 9 will also affect
the product compositions.
S,!!&ry# the potential manipulated variables for product compositions are D,
B, L and 5.
Col,!n 'ontrol tr&tegie :&n intro$,'tion;
' Qbottom - up# approach should be adopted whereby variables that are essential
to operation are regulated before !uality variables. In other words, pressure and
3!
:6.1;
then level must be ade!uately controlled before attention is focused on control of
composition.
Pressure control
4e!uired as a change in pressure will affect relative volatility "U$, the temperature
difference across the reboiler and condenser as well as process safety. '
common pressure control loop is shown below+
Fig 6.+0. A 'o!!on .re,re 'ontrol loo. :PC D .re,re 'ontroller;.
(ere, pressure is regulated using the flow rate of coolant to the condenser.
Increasing or decreasing the water flow rate will alter the temperature of the
condensing li!uid and hence the amount of vapor in the column. This, in turn,
alters the pressure in the column. This will be a slow loop as the dynamics effects
of the cooling can be slow in comparison to simply venting the system by e.g.
opening a valve the ig ..IA also shows this option, which may be re!uired as a
safety mechanism, in case a situation of excessive pressures arose$.
Level control
There will be two level loops on a distillation column as+
The column base level must be maintained at an acceptable value.
The reflux drum level must be maintained at an acceptable value.
The possible schemes that may be employed to do this are summari;ed below+
31

T&%le 6." Poi%le Level Loo. Control
Re&on */y tr&tegy i not .r&'ti'&l#
eed low would not be used to control reflux drum level.
Dottoms flow would not be used to control reflux drum level.
9apor flow would not be used to control reflux drum level.
This scheme violates the mass balance relationships therefore cannot be
used (the reason why will be explained later in the notes).
0istillate flow rate would not be used to control level at the column
base.
32
4eflux flow would not be used to control level at the column base.
rom the matrix of 7C possible alternatives, there are actually only I schemes
that offer acceptable input,output combinations "from a practical viewpoint$.
S'/e!e :I;
Control level in the column base via manipulation of the bottom product
flow rate "by autoatically ad!usting value$.
Control level in the reflux drum by manipulation of distillate flow rate.
' flow controller has been placed on the reflux line (to ensure steady flow
of reflux to the colun)"
Configuring a control strategy+ scheme / "the energy balance control scheme$
Fig 6.+1. S'/e!e :I;6 inventory 'ontrol :PC D .re,re 'ontrol6 LC D level
'ontrol6 FC D )lo* 'ontrol;.
S'/e!e :II;
Control level in the column base by manipulation of the vapor boil,up
through the energy input to the reboiler "in practice this is achieved by
autoatically ad!usting the pressure/ flow of the heating ediu to the
reboiler$.
Control level in the reflux drum by manipulation of distillate flow rate.
33
' flow controller has been placed on the reflux line (to ensure steady flow
of reflux to the colun)"
Configuring a control strategy+ scheme II "a material balance control scheme$
Fig 6.+" S'/e!e :II;6 inventory 'ontrol :PC D .re,re 'ontrol6 LC D level
S'/e!e :III;
Control level in the column base by manipulation of bottoms flowrate.
Control level in the reflux drum by manipulation of reflux flowrate.
34
' flow controller has been placed on the distillate line (to ensure steady
flow of product)"
Configuring a control strategy+ scheme III "a material balance control scheme$
35
Fig 6.++. S'/e!e :III;6 inventory 'ontrol :PC D .re,re 'ontrol6 LC D level
Sele'ting &n &..ro.ri&te $itill&tion 'ol,!n 'ontrol tr&tegy
Dasic Qrules of thumb# can be used to develop feasible strategies. The
methodology "as well as some rules$ is explained below+
R,le o) t/,!% 1# Qflow control the smallest product flow# :& t/i *ill le&ve &
l&rge )lo* tre&! to !&ni.,l&te level;.
E8&!.le 1# S,..oe t/&t t/ere i & l&rge %otto! )lo*r&te :B; &n$ & !&ll
$itill&te )lo* r&te :D;.
>sing this information the control strategy may be developed as follows+
low control the distillate flow "0$
Ensure that the material balance is maintained around the reflux drum.
4ecall that 9 1 0 E 5, for a constant 9, if 0 changes then there must be an
e!ual and opposite change in 5 or the level in the reflux drum will either
drop or start to increase. To ensure that the level remains constant "and
that an appropriate change is made to 5$ a level controller is re!uired on
the reflux drum the manipulated variable being 5.
Ensure that the material balance is maintained around the column base.
4ecall that 1 0 E D so if, for a constant , 0 changes then there must be
an e!ual and opposite change in D or the level in the base of the column
will either drop or start to increase. To ensure that the level remains
constant "and that an appropriate change is made to D$ a level controller
is re!uired with its manipulated variable being D.
This control scheme corresponds to scheme III and is one of the more popular
control schemes. It is often referred to as a material balance control scheme.
E8&!.le "# S,..oe t/&t t/ere i & !&ll %otto! )lo* r&te :B; &n$ & l&rge
$itill&te )lo* r&te :D;.
>sing this information the strategy may be developed as follows+
low control the bottoms flow "D$.
Ensure that the material balance is maintained around the column, 1 0
E D.
or constant , if D changes there must be an e!ual and opposite change in 0 or
li!uid inventory will change "e.g. level may rise in the reflux drum, column base,
or both$. To maintain constant inventory, a level controller is used to make an
appropriate change to 0. @ ensure that the material balance is maintained around
the column base. 4ecall that at the column base E 5 , D 1 9, for a constant
and 5, if D changes then there must be an e!ual and opposite change in 9 or the
level in the base of the column will either drop or increase. To ensure that the
level remains constant "and that an appropriate change is made to 9$ a level
controller is re!uired "the mv being 9$.
This control scheme corresponds to scheme II and it should be noted that the
control of level using 9 may have weird dynamic effects and therefore is not a
favorite. 'gain, this control scheme is often referred to as a material balance
control scheme.
R,le o) t/,!% "# Qmaterial balance control scheme "III$ should be favored if
there is a large reflux ratio, i.e. "5=0$ V C# :i) L i l&rge in comparison to D t/en
rel&tively !&ll '/&nge in L *ill en,re goo$ level 'ontrol6 i.e. t/e .ro'e
g&in i l&rge;.
R,le o) t/,!% +# Qcontrol scheme "I$, often referred to as the energy balance
control scheme, should be favored if there is a small reflux ratio, i.e. "5=0$ W /# :i)
L i !&ll in comparison to D t/en rel&tively !&ll '/&nge in D *ill en,re
goo$ level 'ontrol6 i.e. t/e .ro'e g&in i l&rge;.
Co!.oition 'ontrol
*n,line analy;ers are rarely used as the installed cost will normally be in the
range of X/AA Y per instrument. Therefore composition is often regulated
indirectly using temperature "at constant pressure there is a direct relationship
between temperature and composition for a binary mixture$. >sing a li!uid
temperature near the base of the column for bottom composition and a li!uid
temperature near the top of the column for top product composition, the
remaining mv#s "i.e. those not used for the purposes of level and pressure
control$ may be used to regulate composition. This leads to the following
schemes+
S'/e!e :I;
Top product composition "through a li!uid temperature near the top of the
column$ is regulated by ad6usting reflux flow, 5.
Dottom product composition "through a li!uid temperature near the bottom
of the column$ is regulated by ad6usting vapour flow, 9 "indirectly via
steam flow$.
This gives rise to an alternative name for this control strategy+ t/e L5
'on)ig,r&tion.
Composition Control+ scheme / "the energy balance control scheme$
Fig 6.+0. S'/e!e :I;6 inventory E 'o!.oition 'ontrol :PC D .re,re
'ontrol6 LC D level 'ontrol6 FC D Flo* 'ontrol &n$ TC D te!.er&t,re 'ontrol;.
T/i '/e!e i &lo (no*n & t/e L5 'on)ig,r&tion.
S'/e!e :II;
column$ is regulated by ad6usting reflux flow, 5.
of the column$ is regulated by ad6usting bottoms flow, D.
This gives rise to an alternative name for this control strategy+ t/e LB
'on)ig,r&tion.
Composition Control+ scheme II "a material balance control scheme$
Fig 6.+1 S'/e!e :II;6 inventory E 'o!.oition 'ontrol :PC D .re,re
'ontrol6 LC D level 'ontrol6 FC D Flo* 'ontrol &n$ TC D te!.er&t,re
'ontrol;. T/i '/e!e i &lo (no*n & t/e LB 'on)ig,r&tion.
S'/e!e :III;
column$ is regulated by ad6usting distillate flow, 0.
of the column$ is regulated by ad6usting vapour flow, 9.
This gives rise to an alternative name for this control strategy+ t/e D5
'on)ig,r&tion.
Composition Control+ scheme III "a material balance control scheme$
Fig 6.+6 S'/e!e :III;6 inventory E 'o!.oition 'ontrol :PC D .re,re
'ontrol6 LC D level 'ontrol6 FC D Flo* 'ontrol &n$ TC D te!.er&t,re
'ontrol;. T/i '/e!e i &lo (no*n & t/e D5 'on)ig,r&tion.
@or(e$ e8&!.le# & !et/&nol - *&ter 'ol,!n.
CA = CA wt B methanol = water mixture is to be separated in a /A stage column.
The feed rate is .C,kg=hr entering at stage C. The ob6ective is to separate the
mixture into a top product of HC,wtB methanols and a bottom product of C,wt B.
The feed is li!uid at its boiling point. The condenser is a total condenser. The
reflux flow is I. kg=hr.
a$ What are the materials flows through this system "external li!uid and
internal li!uid and vapor flows$)
b$ %uggest a possible control strategy for this column.
4ules of thumb, common sense and a basic knowledge of chemical engineering
can generally be used to specify an appropriate manipulated variables and hence
the control scheme of a distillation column. (owever, this basic knowledge
should also be complemented by rigorous systems analysis. To do this it is
necessary to consider distillation column modeling in greater detail.

Control: Rolta Academy:Engineering Design Services

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Control: Rolta Academy:Engineering Design Services

Uploaded by

Copyright:

Available Formats

Rolta Academy :Engineering Design Services

Rolta Academy :Engineering Design Services

You might also like