-
Process Control in the Chemical Industries 85
IMPROVEMENTS OF SINGLE-LOOP PID FEEDBACK CONTROL
Introduction
Single loop feedback controllers often provide satisfactory
control performances. Besides their simplicity that reduces
engineering effort, its main advantages are:
It requires minimal knowledge about the process to be
controlled. In particular, a mathematical model is not necessary,
although it is useful for control system design.
The classical PID controllers are versatile and robust. If
process conditions
change, re-tuning the controller usually produces satisfactory
response. Feedback control has also certain inherent
weaknesses:
The feedback controller acts only after the process feels the
upset. It therefore can never yield perfect control where the
controlled variable does not deviate from the set point during load
or set point changes.
Poor feedback tuning may cause instability
PID controller does not always provide the best possible control
for all processes
especially for processes with large dead times and/or cascade
processes with large time constant.
In some applications the controlled variable can not be measured
on line and
consequently feedback control is not feasible.
Feedback control does not provide predictive control action to
compensate for the effects of known or measurable disturbances.
In order to enhance the single loop PID feedback control
performances, a number of
special control configurations such as cascade control,
feed-forward control, and inferential control are often used. Each
of these configurations improves the feedback controller design by
taking advantage of additional knowledge about process dynamics
through one of these means:
Additional process output measurements are used (e.g. cascade
,inferential)
Additional process inputs measurements are used (e.g.
feed-forward)
Use explicit modeling in control calculations (e.g.
inferential)
Use a different control algorithm than PID (e.g.
feed-forward)
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 86
1. Cascade Control Cascade control is one of the most successful
configurations to improve the performances of the single-loop
feedback control. It can provide more effective control by reducing
both the maximum deviations and integral error for the disturbances
response.
As it was explained in the introduction, cascade control uses an
additional measurement of a process variable to assist in the
control task. This technique is shown in the block diagram of Fig.
1. Two controllers are used but only one process variable is
manipulated. The primary (also called outer, master) controller
maintains the primary variable y1 at its set point by adjusting the
set point ysp2 of the secondary controller. The secondary (also
called inner, slave) controller responds both to the output of the
primary controller and to the secondary controlled variable y2
.
GcI GcII GPII GPIy1
Primary Loop
Secondary Loop
+
_+ _
dII+
+ +
+dIy1
sp y2sp y2
Figure 1: Cascade control block diagram
1.1 Illustrating Example
Consider the stirred tank heater of Fig. 2 for which the
objective is to control the exit temperature T using the heating
Fc. For a large disturbance in the heating oil pressure, Figure 3
shows a typical response of the controlled process using a PID
feedback controller. As it can be expected the disturbance is
ultimately suppressed but the response will be slow since the exit
temperature must be disturbed before the feedback controller can
respond.
Feed
Product
Heating oil
T
Fc
Figure 2: Stirred Tank Heater with single control loop
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 87
To improve the speed of the response, we use a secondary
measured process input which is the heating oil flow because it
responds faster to the disturbances in the oil pressure. The
control system (Fig. 4) uses two feedback controllers that can be
standard PID controllers. The output of the exit temperature
controller adjusts the set point of the flow controller. That is
the set point of the secondary controller is equal to the primary
controller output.
Temperature
Flow
Time Figure 3: Dynamic response of stirred tank heater to
disturbance in oil pressure using
single loop controller The secondary flow control loop is
essentially the manipulated variable for the
primary controller. Figure 5 shows a typical response for the
same disturbance using the cascade control. Because the sensor and
valve are fast processes, the flow controller can rapidly achieve
the desired flow of oil. The secondary controller corrects for the
disturbance before the tank exit temperature is significantly
affected by it.
TI
TC
FcHeating oil
Sp1 fromoperator
F1
P0
P1
F0T0
Figure 4: Stirred Tank heater with cascade control
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 88
Temperature
Flow
Time Figure 5: Dynamic response of stirred tank heater to
disturbance in oil pressure with
cascade control
1.2 Advantages The advantages of cascade control can be
summarized as follow:
Disturbances felt by the secondary variable is significantly
corrected by the secondary controller before it is felt by the
process.
The dynamics of the secondary loop are much faster than those of
the primary
loop. The cross over of the secondary loop is higher than that
of the primary loop. This allows the use of higher gains in the
secondary controller to suppress more effectively the effect of the
disturbance occurring in the secondary loop without affecting the
stability of the system.
1.3 Tuning and Implementation Issues 1.3.1 Selection of the
secondary variable The key point in cascade control is the
selection of secondary variable. Two guidelines must be
observed:
The secondary variable must indicate the occurrence of an
important disturbance
The secondary variable dynamics must be faster that the primary
variable dynamics
1.4 Types of Cascade Controllers
Cascade control can use the standard PID feedback controllers in
the two loops. The secondary loop must have the proportional mode
but it does not require the reset mode. Integral model may be used
in the secondary controller if it is desired to suppress completely
the disturbance entering the primary or when the primary controller
is not in operation (sensor not functioning or calibrated, etc).
Derivative modes are not advised in the secondary loop since the
derivative action is designed to overcome some lag in the
controller loop and if applied to set point changes may result in
excessive valve motion and overshoot.
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 89
The cascade controller is tuned in a sequential manner. The
secondary controller is first tuned satisfactorily before the
primary is tuned . Conventional tuning guidelines for PID apply for
both the control loops. 1.5 Instability and Saturation
Adding cascade control to a process can destabilize the primary
loop if most of the process dynamics are within the secondary loop.
A common example is the use of a valve positioner in a flow control
loop. Closing the loop around the valve increases its gain so that
the proportional band of the flow controller has to be increased to
maintain stability. This leads to slower response. If large valve
motor or long pneumatic transmission lines cause problems a volume
booster should be used to load the valve motor rather than a
positioner.
Saturation problems can arise when both primary and secondary
controllers have automatic reset. The need for anti-reset windup is
much greater for cascade designs since once the secondary loop is
saturated, the primary controller will also saturate. When putting
a cascade system into automatic operation the secondary controller
must be first transferred to automatic. 1.6 Examples of Commonly
Used Secondary Loops 1.6.1 Valve positioner
Stem friction and/or changes in line pressure can cause a
hysteresis between the action of the control signal and its effect
on the valve position. i.e. The valve may remain stationary and
then jump to a value beyond that necessary to bring the controlled
variable to its set point. This can degrade performances.
The effects of a valve sticking can be reduced by the use of a
valve positioner
which is included as part of valve equipment (Fig. 6).The
primary controller sends it signal to the valve positioner which in
turn adjusts the air pressure until the desired stem position is
nearly achieved. Valve positioner are proportional controllers with
smaller proportional band. This allows for fast dynamics that will
reduce the dead zone in the control valve and improve the response
of the valve.
Vc
Tc
Remote location
positioner locatedat valve
Figure 6: Schematic of a valve positioner
1.6.2 Flow control To ensure that line pressure fluctuations or
undesirable valve characteristics do
not affect the primary loop, flow control can be set in cascade
as it is the case in composition control. Flow controllers may have
proportional and reset modes.
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 90
1.6.3 Temperature Control
When there is need for accurate control of heat transfer such as
in jacketed continuous stirred tank reactor (Fig. 7), cascade
control is used. The reactor temperature is controlled by
manipulating the coolant temperature in cascade.
TT
Tc
TT
set point
Fccoolant Tc
Feed
Product
BA
TT
Tcset point
Fccoolant Tc
Feed
Product
BA
set point
Figure 7: Temperature control: (a) conventional feedback; (b)
Cascade control
1.7 Illustrating Example of Design of Cascade Control
Consider the block diagram of a cascade system shown in Fig 1.
To simplify the presentation we assume that the transfer functions
of the measuring devices is one. The dynamics of the secondary loop
are:
Gsecondary = GcII(s)GpI(s) (1) Figure 8 shows a simplified form
of the general block diagram where the secondary loop has been
considered as a dynamic element. The overall transfer function for
the primary loop is
IIIII
IIIIIprimary GpGpGc
GpGcGcG )
1(
+=
(2)
The stability of the primary loop is determined by the
characteristic equation,
IIIII
IIIIIprimary GpGpGc
GpGcGcG )1
(11+
+=+ (3)
Let see how the design can be carried out for the following
example:
))(.( 11501
++=
ssGpI
and
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 91
).( 1101
+=
sGpII
Note that from looking at the time constants that the secondary
process is faster than the primary.
GcI GPI y1y1
sp
+_
dII
+
+
+dI
PIIcII
PIIcIIGG
GG+1
PIIcIIGG+11
+
Figure 8: Simplified cascade control block diagram
Simple feedback control Using a simple PI controller with reset
time I = 1, the open loop transfer function is
+++
+=
)1)(15.0)(11.0(111
ssssKGpGpGc cIIIII
(4)
The cross over frequency can be found from the equation that
sets the total phase lag to 1800
( ) ( ) ( ) 180tan5.0tan1.0tan1tan 1111 =+++
cococo
co
wwww
(5)
This yields:
wco = 4.45 rad/.min The overall amplitude ratio is given by
1
1
150
1
110
1112222 +++
+=wwww
KcAR I).().(
(6)
The ultimate value of the gain KcI can be found from the
condition
AR = 1 at w = wco Thus
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 92
1
1
150
1
110
11112222 +++
+=cocococo
Iwwww
Kc).().(
(7)
yielding
KcI = 11.88 Therefore when the disturbance dII of the secondary
process changes, the simple feedback controller can use a gain up
to 11.88 before the system becomes unstable. Also because the
overall process is of the third order we expect that the closed
loop response of y(t) to change in dII will be rather sluggish.
Cascade Control
Consider a cascade control system similar to that of Fig. 1. The
open loop transfer function for the secondary loop is given by Eq.
1. Assuming a simple proportional controller yield:
11.01
+=
sKcGpGc IIIII
There is no cross over frequency for the secondary control loop.
Large values for the gain KcII can be used that yield fast closed
loop responses. Once KcII is selected for the secondary loop, the
cross over frequency for the overall process can be obtained as
before. Then KcI can be selected for the primary controller using
Ziegler-Nichols method. 2. Feed-forward Control
As mentioned in the introduction, feed forward control attempts
to enhance the
performance of the single loop feedback control by making use of
an additional measurement of process input. Figure 9 shows the
general form of feed-forward control configuration. The disturbance
is measured directly and the manipulated variable is changed
accordingly to eliminate the impact of the disturbance on process
output. Recall that feedback controller reacts only after it has
detected a deviation of the value of the output from the desired
steady state, therefore feedback control can never achieve perfect
control. Feed-forward control on the other hand anticipates the
effect the disturbance will have on the process output. Therefore
feed forward controllers have the potential, at least theoretical,
for perfect control.
In many applications, feed flow is the primary component of load
because it can
change widely and rapidly. Load elements can also be feed
composition when the product composition is to be controlled, or
feed flow and temperature when the product temperature is to be
controlled.
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 93
Process
Controller Disturbance
Manipulatedvariable Controlled variable
Figure 9: Feed-forward controller 2.1 Examples
The objective of the feed-forward controller is to compute the
manipulated variable that would balance the load (load balancing).
One may anticipate that the logic of the feed-forward will depend
on the availability of mathematical equations relating the load to
the manipulated variable. These equations are the material and
energy balance either at steady state or dynamic.
Consider the example of the stirred tank heater of Fig 10. The
objective is to
design a feedforward control for the temperature by measuring
the disturbance Ti and using the amount of heat Q provided by the
steam as manipulated variable. Since the focus will be only on
controlling the temperature we assume that the level L is constant
(i.e. Fi does not change).
Ti, Fi
Steam
T, F
L
Figure 10: Stirred Tank Heater
The process variables are related by the dynamic heat balance
equation:
CpQTTF
dtdTAL ii
+= )( (8)
2.1.1 Steady state feed-forward controller
The simplest form of feed-forward control is the one based on
steady state balance, Setting dT/dt = 0 in Eq. 8, yields:
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 94
)( ii TTCpFQ = (9) Therefore in order to keep the controlled
variable, i.e. temperature T at the set point Tsp, the manipulated
variable i.e. amount of heat should be:
)( isp
i TTCpFQ = (10) This simple relation is the design equation for
the static feed-forward controller.
Figure [11a] shows the block diagram for this static
feed-forward control. Figure [12a] shows the temperature response
to a load change in feed temperature.
CpFi
+
_
1 +sTsp
Ti
Ti, Fi
T, FSteam
Controller
Feedforwardcontrol
CpFi
+
_
1Tsp
Ti
Ti, Fi
Steam
Controller
Feedforwardcontrol
(a)
(b)
T, F
Figure 11: feed-forward of temperature control; (a) static; (b)
dynamic
2.1.2Dynamic Feed-forward controller
To improve the controller transient response, we will use the
dynamic heat balance to design a dynamic feed-forward controller.
Equation (8) can be also put in the following form
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 95
CpFQTT
dtdT
FV
ii
i +=+
(11)
where V=AL is the constant liquid volume in the tank. The
equation can also be put in deviation variables form as
follows:
CpFQTT
dtdT
FV
ii
i +=+
'''' (12)
Taking Laplace transform yields:
CpFsQsTsTssT
FV
ii
i +=+
)(')(')(')(' (13)
Or equivalently:
CpFsQ
sssTsT
i
i
++
+=
)(')(')('1
11
(14)
The objective is to maintain T' at T'sp i.e. T''(s)=T''sp(s),
therefore Equation 14 yields:
)](')(')[()(' sTsTsCpFsQ isp
i += 1 (15)
where = V/Fi is the retention time of the liquid in the tank.
Equation (15) is the design equation for the dynamic feed forward
controller.
For load changes, the steady state and dynamic controllers will
be equivalent but dynamic feed-forward will be better for set-point
changes as shown in Fig 12b.
time t = 0
TDynamic F.F.
Static F.F.
Response to setpoint step changeResponse to load step
changestatic = dynamic F.F.
(a) (b) Figure 12: Response to load response and set point
change using feed-forward control
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 96
2.2 Design of Feed-forward Controllers The example before can be
generalized for the design of general static and
dynamic feed forward controllers. Consider the block diagram of
the process shown in Figure 13, it is easy to see that:
( ) )()()()()()()()()()()()( sdsGsGsGsGsGsysGsGsGsGsy mcpdspspcp
ff += (16) Equation (16) determines the form of the feed-forward
control and also the two transfer functions for the design of the
control mechanism:
Gc
GPy
ysp +_
+
+
d
Gd
Gf
Gm
Process
Final controlelement
Measuring device
Feedforward controlmechanism
Gsp
Figure 13: General block diagram for feed-forward control
mechanism For Load suppression, the coefficient in d(s) in equation
(16) should be zero. The design equation is therefore
)()()()()(
sGsGsGsGsG
mp
dc
f
= (17)
Similarly, for set point change, the design equation, setting
the coefficient of ysp in equation (16) to zero is:
)()()(
sGsGsG
d
msp =
(18)
Figure 14 shows the feed forward control configuration with the
control mechanisms. These design equations illustrate some of the
basic characteristics of feed forward control.
From Eqs (17,18), it is clear that feed forward controllers can
not be conventional PID controller. Instead they should be viewed
as special-purpose computing machines.
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 97
The design equations (17,18) illustrate that feed forward
control depends heavily on the good knowledge of process elements
Gp, Gd. Poor knowledge of any of these processes deteriorates the
performance of feedforward control and prevent the achievement of
perfect control.
GPy
ysp +_
+
+
d
Gd
Gf
Gm
Process
Final controlelement
Measuring device
Feedforward controlmechanism
d
msp G
GG =
mp
dc GGG
GGf
=
Figure 14: block diagram for feed-forward control mechanism
2.2.1 Design of steady state feedforwad elements
To illustrate the design procedure for the stirred tank heater,
we will assume that Gm(s) = Gf(s) = 1. From Equation (14) we can
identify the process Gp(s) and Gd(s):
11
1
1
+=
+=
sG
sCpFsG dip ;)(
At the steady state only the static elements of the process
transfer functions of Gp(s) and Gd(s) are retained; i.e. Gp = 1/Fi
Cp Kp; Gd =1 Kd. The design transfer functions (Equations 17 &
18) are reduced in this case to the simple constants:
dsp
p
dc
KsG
KKsG
1=
=
)(
)(
(19) (20)
2.2.2 Design of dynamic feedforwad elements
If the process model Gp(s) and Gd(s) are known we can proceed
with using Equations (17,18). If they are not known, it is possible
to use approximations to them and still obtain good results over
the steady state feed-forward controller. Assume that Gp(s) and
Gd(s) are approximated by first order lags:
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 98
11
11
+=
+=
ssG
ssG
d
p
)(
)(
(21)
(22)
That is should be equal to the time constant of the controlled
variable in
response to the manipulated variable and should be the time
constant of the controlled variable in response to the load. The
design equations (Eqs. 17,18) yield:
11
11
11
++
=
+
+==ss
s
ssGsGsG
p
dc )(
)()(
(23)
and
11 +== ssG
sGd
sp )()(
(24)
The transfer function of Gsp(s) is called a lead element while
that of Gc(s) is a
lead-lag. (i.e. 1/lead lag).The term s+1 introduces phase lead.
It tends to improve the rise time and overshoot of the system
response but increase the bandwidth. Phase lag introduced by 1/(
s+1) improves the steady state response (stability margin) but it
results in longer rise time because of the reduced bandwidth.
Lead-lag elements are expected therefore to produce satisfactory
performance and compensate each other. Lead-lag elements are the
most commonly used elements in dynamic feed forward control. Figure
15 shows the response of lead-lag unit to a unit step.
0 2 4 6 8Time (min)
100.0
0.5
1.0
1.5
2.0
Res
pons
e
= 1 min
= 2.01.51.0
0.75
0.0
Figure 15: Unit step response of lead-lag unit
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 99
2.3 Tuning of the lead-lag elements The adjustable parameters (,
) of the lead-lag can be selected following this procedure:
1.With no dynamic compensation (lead and lag time constant sets
at equal values), Introduce a load change to observe the direction
of the error. If the resulting response is in the direction of the
load, the lead time should exceed the lag time, i.e. > . It can
be seen from Figure 15. that in this case the compensation must
over correct at first since the controlled variable responds faster
to the disturbance than to the manipulated variable. If the
response is not in the direction of the load then the lag time
should be grater >. It can be seen in Fig. 15, that in this case
the compensator must delay its action in order to prevent
correcting too soon since the controlled variable responds faster
to the manipulated variable than to the disturbance.
2.Next measure the time required for the controlled variable to
reach its
maximum or its minimum value. Set this time to be the smallest
of the two time constants. If the lead time is larger this time
should be the lag time and vice versa.
3.Set the greater time constant at twice this value and repeat
the load change. If
the error curve is still not equally distributed across the set
point one should increase the greater time constant and repeat the
load change.
4.Once the error area is equally distributed about the set point
both settings
should be increased or decreased with their difference constant
until a minimum error amplitude is achieved.
Finally, since feed-forward control is more costly and requires
more engineering effort, its use is restricted therefore to
applications deemed justifiable. Most feedforwad control systems
have been applied to processes that are very sensitive to
disturbances and slow to respond. Examples include:
Distillation columns
Boilers
Multiple Effect evaporator Solid dryers
3. Feed-forward Feedback Control
Despite its advantages, feedforward control suffers from the
following inherent weaknesses:
It requires the identification of all possible disturbances and
their direct measurements something that may not be possible for
many processes
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 100
Any changes in the parameters of the process can not be
compensated by a feedforwad controller since their impact can not
be detected.
Feed forward control requires a very good model for the process
which is not
possible for many processes.
Feedback control on the other hand is insensitive to all three
drawback but has poor performance for a number of systems
(multicapacity, dead time), and raises question of closed loop
stability. We expect that a combination of the two controller would
yield better performance
Figure 16 shows how a feedback control can be added to the
feedforwad loop.
The open loop transfer function is:
y(s) = Gp(s)m(s) + Gd(s)d(s) (25) The value of the manipulated
variable is given by:
m(s) = Gf(s)c(s) = Gf(s) [c1(s)+c2(s)] =
Gf(s)Gc1(s)e1(s)+Gf(s)Gc2e2(s) (26) or
m(s) = Gf(s)Gc1(s)[ysp(s)-Gm1y(s)] +
Gf(s)Gc2[Gsp(s)ysp(s)-Gm2(s)] (27) Recasting the expression of m
from Eq. 25 yields:
( )
)()()()()(
)()()()()()(
)()()()()()()()()(
)( sdsGsGsGsG
sGsGsGsGsGsy
sGsGsGsGsGsGsGsGsG
symcp
mcpdsp
mcp
spccp
11f
22f
11f
21f
11 +
++
+=
(28)
Note that the stability of the closed loop response is
determined by the roots of the characteristic equation:
1+Gp(s)Gf(s)Gc1(s)Gm1(s) = 0
which depends on the transfer function of the feedback loop only
and does not depend on the addition of the feedforwad loop. The
transfer functions of the feedforwad loop Gc2 and Gsp are given by
(Eqs. 17,18).
d
msp
mp
dc
GGG
GGGGG
2
2f2
=
=
(28) (29)
If Gp, Gd, Gf and Gm2 are known only approximately, then:
Gd Gp Gf Gc2 Gm2 0 and Gp Gf Gc2 Gsp 1
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 101
In such case the feedforwad does not provide perfect control y
ysp ( e1 0) and the feedback is activated to offer the necessary
compensation.
Gc1 Gf Gp
Gm1
Gc2
Gsp Gm2
Gd
y(s)
d(s)
m(s)c(s)c1(s)
c2(s)
e2(s)
e1(s)ysp(s)
ysp(s)
Feedbackcontroller
Feedforwardcontroller
Sensor meeasuringdisturbance for feedforward
controller
Sensor meeasuring output forfeedback controller
+
++
++
+ _
_
Figure 16: Generalized block diagram for feed-forward feedback
controller
3.1 Feed-forward-feedback Control of a Tank Heater
The example for a tank heater with a feed-forward and PI
controller is shown in Figure 17. The design equations for feed
forward control alone are: Gc2 = Fi Cp and Gsp = s+1.
)(s
KG cc1
111
+=
CpFGc 12 =++
+
_e
+
_
1+= sGsp Tsp
Ti
Ti, Fi
T, F
Figure 17: Example of heated tank with feed-forward-feedback
controller
Assume that the density or heat capacity Cp are not known
exactly, then the
feed-forward control alone does not provide perfect control.
Figure 18 shows the temperature in the tank after a step change in
the inlet temperature. Because of the uncertainty in there is
offset in the response. With feedback PI control added to the feed
forward loop and for the same step change, the deviation has
disappeared.
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 102
0time
T
Feedforward control
Feedforward-Feedbackcontrol
Deviation remaining fromfeedforward control only
Figure 18: Temperature response
4. Ratio Control
In some aspects ratio control can be considered as a special
type of feed-forward
control where two loads are measured and held in a constant
ratio to each other. It is mostly used to control the ratio of flow
rates of two streams. Both flow rates are measured but only one can
be controlled. The stream whose flow rate is not under control is
refereed to as wild stream. Considering Figure 19, there are
essentially two configurations for ratio controller:
In configuration 1 we measure both flow rates and take their
ratio, This ratio is compared to the desired ratio (set point) and
the deviation (error) between the measured and desired ratio
constitutes the actuating signal for the ratio-controller
In configuration 2 we measure the flow rate of the wild stream
and multiply it
by the desired ratio. The result is the flow rate that the
stream B should have and constitutes the set point value which is
compared to the measured flow rate of stream B. The deviation
constitutes the actuating signal for the controller which adjusts
appropriately the flow of B.
Ratio control is used for a variety of applications
including:
Keep constant the ratio between the feed flow rate and the steam
in the reboiler of a distillation column,
Hold constant the reflux ratio in a distillation column.
Control the ratio of two reactants entering a reactor at a
desired value.
Hold the ratio of two blended streams constant in order to
maintain the
composition of the blend at the desired value.
Hold the ratio of a purge stream to the recycle stream
constant.
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 103
Keep the ratio of fuel/air in a burner at its optimum value
Maintain the ratio of the liquid flow rate to vapor flow rate in
an absorption
constant.
FT
Divider
FT
GR
FA
ValveB
A
FB
MeasuredRatio
Desired Ratio
Error
+_
Wild stream
Controlable stream
Ratiocontroller
FT
FT GR
FA
ValveB
A
FB
DesiredRatio
Error
+_
Wild stream
Controlable stream
Ratiocontroller
Figure 19: Ratio controller configuration
5. Selective/Override Control
Most process control problems have an equal number of controlled
variables and
manipulated variables. If a process has fewer manipulated
variables than controlled variables, a strategy is needed for
sharing the manipulated variables among the controlled variables. A
common strategy is to use selectors to choose the appropriate
process variables among a number of available measurements. Signal
selectors choose either the lowest, median or the highest control
signal from among two or more signals. A control loop containing
this type of logic is called selective control. In this section we
present examples of the use of this type of control.
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 104
5.1 Maintaining Safety of the Equipment
When a process variable exceeds certain given hazardous limits,
the equipment is automatically shut down by process interlocks.
Shutdowns can be avoided by the use of secondary controller that
has a higher priority and overrides the normal control operation,
thus keeping the process running at some suboptimal conditions.
Examples of these situations include:
Safeguard the operation of variable speed pumps.
Safeguard the operation of high temperature or pressure
reactors.
Avoid flooding in distillation columns
Safeguard the operation of furnace. Figure 20 illustrate the
concept of selective/override control in a pumping system
for a sand-water slurry. During normal operation the level
controller (LC) adjusts the slurry exit flow by changing the pump
speed. The slurry velocity in the exit line must be however always
kept above a minimum value to prevent the line from sanding up. If
the flow rate (and hence the velocity) decreases and approaches a
lower limit the flow controller takes over from the level
controller and speeds up the pump. The strategy is implemented
using a high selector and a reverse acting flow controller with a
high gain. The selector compares signals P1 and P2 and chooses the
highest one. This type of control is also called override
control.
FT
FCHS
LCLT
h
Slurry in
Holding Tank
hm
p2
p1
pqm
qSlurry out
Figure 20: A selective control for sand-water slurry system
5.2 Improving Control Performance
A plug flow reactor where an exothermic reaction is taking place
is always prone to the developments of hot spots. For good control,
the sensor providing the temperature measurement should be located
at the hot spot. As the catalyst in the reactor ages or conditions
in the reactor change, the hot spot move along the reactor. It is
desired to design control scheme so that the measured variable
"moves" with the hot spot. A control strategy that accomplishes
this goal is shown in Fig 21. The high
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 105
selector selects the transmitter (all transmitters assumed to
have the same range) with the highest output and the control is
based on this temperature.
TT TC
SP
Reactants
Cooling meduim
Products
TTTTTT
HS
Figure 21: A plug flow reactor
5.3 Optimization of the process
Consider the furnace of Fig 22. where fuel oil is used to
provide heat to a number of process units. Each individual unit
manipulates the flow of oil required to maintain its controlled
variable at set point. A bypass control loop is also provided. A
bad or inefficient operation of the process is the one for which
the oil temperature is heated above the value that would satisfy
the need of the users. In this case most of the valves would not be
wide open and large quantity of fuel would be burned to reach the
unnecessary high oil temperature.
Fuel spsp
sp
sp
sp
Hot oil
Stack gas
T
T
T
h
Figure 22: Hot oil system
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 106
The effective operation that would save energy is the one that
would maintain the oil leaving the furnace at a temperature just
enough to provide the necessary energy to the users with hardly any
flow through the bypass valve. In this case most of the temperature
control valves would be open most of the time. To achieve this
goal, the selective control strategy, shown in Fig. 22, first
selects the most open valve using a high selector .The valve
position controller controls the selected valve position at large
value i.e. 90 % open by manipulating the set point of the furnace
temperature. This saves energy because it will maintain the
temperature just hot enough to provide needed heat to the users.
5.4 Protecting against sensor/transmitters failures
Selectors are also used to protect against transmitter failures
by selecting a valid transmitter signal among several. Redundant
transmitters monitor the process variable and the median selector
chooses the right one for control. Redundant sensors are commonly
used in a hostile environment of high temperature or corrosive
where failures rate are high thus avoiding the shutdown of the
process.
6. Inferential Control As mentioned in the introduction,
inferential control makes use of a secondary
measurement. Quite often the controlled output of a process unit
can not be measured directly and also the disturbance is
unmeasured. This is the type of control where feedback and feed
forward can not be used. Inferential control is the only solution.
Consider the block diagram of the process shown in Fig 23, with one
unmeasured controlled output y and one secondary measured output z.
The manipulated variable m and the disturbance d affect both
outputs. The disturbance is assumed to be unmeasured. The open loop
transfer function in the block diagram of Fig. 23 is:
y(s) = Gp1(s) m(s) + Gd1(s) d(s) (31)
z(s) = Gp2(s) m(s) + Gd2(s) d(s) (32) We can solve for d(s) in
the second equation to find the following estimate of the
unmeasured disturbance
)()()(
)()(
)( smsGsG
szsG
sdd
p
d 2
21 = (33)
Substituting back in equation (31) yields,
)()()()(
)()()()( sz
sGsGsmG
sGsGsGsy
d
dp
d
dp
2
12
2
11 +
=
(34)
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 107
GP2
ym
+
++
d
Process
Controlled output(unmeasured)
Manipulatedvariable
Gd1
GP1
Gd2
+
z
Unmeasureddisturbance
Secondarymeasurement
Figure 23 Process with need for inferential control
This equation provides the estimator needed which relates the
unmeasured
controlled output to measured variables m(s) and z(s). Figure 24
shows the resulting block diagram for the inferential control.
Equation (34) also illustrates the basic characteristics of the
inferential control: The accuracy of the control scheme depends on
the good estimation. This is turn depends on the good knowledge of
the process i.e. Gp1(s), Gp2(s), Gd1(s) and Gd2(s). Generally these
process elements are not known perfectly and therefore the
inferential control would provide control with varying quality.
Generally inferential control is used when composition is the
desired controlled
variable. Temperature in the most common secondary measurement.
Examples of these situations include: chemical reactor,
distillation columns, driers, absorber, etc.
ysp m
+
+_
Estimator
Set point
+
z
Controller
Estimates of unmeasuredoutput y
Gc Process
y
221
1 pGdGdG
pG =
21
dGdG
Figure 24: Process under inferential control system
Example: Inferential control of a distillation column
Consider a distillation column which separates a mixture of
propane-butane in two products. The feed composition is the
unmeasured disturbance and the control objective is to maintain the
overhead product molar composition 95% propane. The
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 108
reflux ratio is the manipulated variable. The feed and overhead
composition are unmeasured so there is need for inferential
control. The secondary measurement to infer the overhead
composition is the temperature at the top tray. The process inputs
are the feed composition (disturbance)and reflux ratio (manipulated
variable) while the outputs are the overhead propane composition
(unmeasured controlled variable) and temperature of top tray
(secondary measurement). The transfer functions (Figure 25a)
between the process inputs and outputs are given in the block
diagram. Following the procedure established before we derive the
following input-output relations,
)(.)(.)( smsesd
sesy
ss
13021
17090 2
++
+=
(35)
)()(.)(.
sms
sdsesz
s
1201
16020 20
++
+=
(36)
The resulting block diagram can be shown in Figure 25b.
ym
+
++
d
Process
Overheadcomposition
Manipulatedvariable
+
z
Unmeasureddisturbance
Temperatureof top tray
130
1211 +
=
s
ssepG
.
1201
2 +=
spG
170
2901 +
=
s
ssedG
.160
2202 +
=
s
ssedG
.
(a)
ysp m
+
+_
Estimator
Set point
+
z
Controller
Estimates of unmeasuredoutput y
Gc Process
y
17016054
++
ss.
))(()(..
12017016054
13021
+++
+
sss
s
se
(b) Figure 25: (a) Block diagram of distillation column; (b)
corresponding inferential
control system
Chemical Engineering Department King Saud University, 2002
-
Process Control in the Chemical Industries 109
References Marlin, T., Process Control: Designing Processes and
Control Systems for Dynamic Performance, McGraw Hill, New York,
1995. Seborg, D., Edgar, T., and Mellichamp, D., Process Dynamics
and Control, Wiley & sons, New York, 1989. Stephanopoulos, G.,
Chemical Process Control: An Introduction to Theory and Practice,
Prentice Hall, 1984. Smith, C. and Corripio, A., Principles and
Practice of Automatic Process Control, Wiley & sons, New York,
1997. Shinsky, F., Process Control Systems, McGraw Hill, New York,
1988. Jones, B., Instrumentation, Measurement and Feedback, McGraw
Hill, New York, 1977. Wightman, E., Instrumentation in Process
Control, Butterworth, 1972. Ogunnaike, B. and Ray, W., Process
Dynamics, Modeling and Control, Oxford University Press, UK, 1994.
Murrill, P., Application Concept of Process Control, ISA, NC, USA,
1988. Murrill, P., Fundamentals of Process Control Theory, ISA, NC,
USA, 1991. Luyben, W.,Process Modeling, Simulation and Control for
Chemical Engineers, McGraw Hill, New York, 1990.
Chemical Engineering Department King Saud University, 2002