Module 4 : Control-Loop Characteristics :

Chapter provides process-control technologists a

general background in the practical considerations of
process-control loop implementations.
CONTROL SYSTEM CONFIGURATIONS :
Single Variable : 1) Independent Single Variable
2) Interactive Single Variable
3) Compound Variable
Cascade Control :

MULTIVARIABLE CONTROL SYSTEMS :

Single Variable :
• The elementary process-control loop is a single-
variable loop. The loop is designed to maintain
control of a given process variable by manipulation
of a controlling variable, regardless of the other
process parameters.
Independent Single Variable :
• In many process-control applications, certain
regulations are required regardless of other
parameters in the process. In these cases ,a setpoint
is established, controller action is started, and the
system is left alone. Refer figure-1
• In Figure 1, a flow-control system is used to
regulate flow into a tank at a fixed rate determined
by the setpoint.
• This system then makes adjustments in valve
positions as necessary following a load change to
maintain flow rate at the setpoint value.
Interactive Single Variable :
• A second single-variable control loop, also shown
in Figure 1, regulates the temperature of liquid in
the tank by adjustment of heat input. This is also a
single-variable loop that maintains the liquid
temperature at the setpoint value.
• Under nominal conditions, the flow into the tank is
held constant and the temperature is also held
constant, both at their respective setpoint values.
• However a change in the setpoint of the flow-
control system appears as a load change to the
temperature control system, because the fluid level
in the tank or rate of passage through the tank must
change.
• The temperature system now responds by resetting
the heat flux to accommodate the new load and
bring the temperature back to the setpoint.
• We say, then, that these two loops interact.
• Almost any process where several variables are
under control shows such interactive behaviour.

IMPORTANT : Any cycling or other instability of the

flow-control loop causes cycling in the
temperature system because of this interaction.
Compound Variable :
• In some cases, a single process-control loop is used
to provide control of the relationship between two
or more variables.
• This can be accomplished by using measurements
from, say, two sensors as input to the process
controller.
• A signal conditioning system must scale the two
measurements and add them prior to input to the
controller for evaluation and action.
• For example refer figure 2
• A common example is when the ratio of two
reactants must be controlled.
• In this case, one of the flow rates is measured but
allowed to float (that is, not regulated),
• The other is both measured and adjusted to
provide the specified constant ratio.
• An example of this system is shown in Figure 2.
The flow rate of reactant A is measured and
added, with appropriate scaling, to the
measurement of flow rate B.
• The controller reacts to the resulting input signal
by adjustment of the control valve in the reactant
B input line.
Cascade Control :
• The inherent interaction that occurs between two
control systems in many applications is sometimes
used to provide better overall control.
• One method of accomplishing this is for the
setpoint in one control loop to be determined by
the measurement of a different variable for which
the interaction exists.
• A block diagram of such a system is shown in Figure
4.
• Two measurements are taken from the system and
each is used in its own control loop.
• In the outer loop, however, the controller output is
the setpoint of the inner loop.
• If the outer loop controlled variable changes, the
error signal that is input to the controller effects a
change in setpoint of the inner loop. Even though
the measured value of the inner loop has not
changed,
• The inner loop experiences an error signal, and
thus new output by virtue of the setpoint change.
• Cascade control provides better control of the
outer loop variable than is accomplished through
a single variable system.
• An example of a cascade control shows how it
works, also suggests how control is improved.
• Consider the problem of controlling the level of
liquid in a tank through regulation of the input flow
rate.
• A single-variable system to accomplish this is shown
in Figure 5a. A level measurement is used to adjust a
flow-control valve as a final control element.
• The setpoint to the controller establishes the desired
level. In this system, upstream load changes cause
changes in flow rate that result in level changes.
• The level change is, however, a second-stage effect
here. Consequently, the system cannot respond until
the level has actually been changed by the flow rate
change.
• Figure 5b shows the same control problem solved
by a cascade system.
• The flow loop is a single-variable system as
described earlier, but the setpoint is determined
by a measurement of level.
• Upstream load changes are never seen in the
level of liquid in the tank because the flow-control
system regulates such changes before they
appear as substantial changes in level.
 MULTIVARIABLE CONTROL SYSTEMS
• A reasonably complex industrial process is
multivariable because many variables exist in the
process and must be regulated.
• In general, many of these are either noninteracting
or the interaction is not a serious problem in
maintaining the desired control functions.
• In such cases, either single-variable controls or
cascade loops suffice to effect satisfactory control of
the overall process.
• The use of the word multivariable refers to those
processes wherein many strongly interacting
variables are involved.
• Such a multivariable system can have a complex
interaction pattern that the adjustment of a
single setpoint causes a profound influence on
many other control loops in the process.
• In some cases, instabilities, cycling, or even
runaway result from the indiscriminate
adjustment of a few setpoints.
 CONTROL SYSTEM QUALITY
• When a manufacturing concept is to be implemented,
the ultimate goal is to develop a product that satisfies
“design” criteria.
• Example 1)
If the product is crackers, they should have a
certain colour, flavour, salinity, size, and so forth.
• Example 2)
If it is gasoline, it should possess certain
octane, antioxidants, viscosity, and so forth.
• The manufacturing process depends on the operation of
a set of process-control loops to impart the desired
characteristics to the product.
• The ultimate gauge of control system quality is,
therefore, whether such control provides a product
that is within specifications.
• In brief, given that a control system can provide a
product that meets specifications, then ask how
well it does perform this job, what variation in
parameters exists,
• Therefore we must first describe measures of
quality in a control system and then analyze how
the loop characteristics affect these measures.
Definition of Quality :
• It is impossible for a control loop to regulate this
variable to exactly the setpoint.
• Let’s face it, the variable must change before the
loop can generate a corrective action to oppose
the change.
• Then, in considering quality, we must accept that
perfect control is impossible. and some inevitable
deviations of variables from the optimum values
will occur.
• “Definition of quality is concerned with these
deviations and their interpretation in terms of the
ultimate product “.
• a set of measures or criteria has been devised.
• These criteria provide a common language so that a
product can be evaluated in terms of the dynamic
characteristics of a specific loop.
• Therefore measure of quality.
• To understand the measure, we must first define
quality in terms of the process control loop.
Loop Disturbance :
The process-control system is supposed to
provide regulation so that disturbances in the system
will cause minimum deviation of the controlled
variable from the setpoint value.
• The “ quality of the control system is defined by
the degree to which the deviations that result from
the disturbances are minimized ” .
• There are three basic types of disturbances that can
occur in a process-control system:
1. Transient
2. Setpoint changes
Transient :
• A transient disturbance results from a temporary
change of some parameter in the system that
affects the controlled variable.
• It is impractical to use transient disturbances to
define control quality because the nature of a
transient cannot be well defined.
• That is, the transient can vary in duration, peak
amplitude, and shape.
• Therefore for definition of quality, we need a more
regular type of disturbance.
Setpoint changes :(permanent change)
• A step-function change in setpoint, as
• shown in Figure 7a, is an instantaneous change of
the loop setpoint from an old value to a new
value.
Load changes :
• The second possible disturbance is a step-function
change in process load, as shown in Figure 7b,
that also occurs instantaneously in time.
• The load change can come from the sudden
permanent change of any of the process
parameters that constitute the process load.
 To provide measures of quality,
we evaluate
how the system responds
to either of these sudden changes .
(That is setpoint change and Load
change)
Optimum Control :
The most universal definition of quality in a
control system is
“ The system provides optimum control—that
is, the best control possible “.
If anything is modified in the system, then the
deviation of controlled variable from a load or
setpoint change is always worse.
Thus, the overall settings of the system are at an
optimum.
This does not mean that the control is “perfect” or
even very good; it simply means that it is the best it
can be.
 Optimum control, and therefore control
quality, can be defined in terms of the three effects
resulting from a load or setpoint change:
1. Stability
2. Minimum deviation
3. Minimum duration
Stability :
The most basic characteristic in
defining process-loop quality is that it provides
stable regulation of the dynamic variable.
• Stable regulation means that the dynamic variable
does not grow without limit.
• In Figure 8, two types of unstable responses are
shown.
• In one case, a disturbance causes the dynamic
variable simply to increase without limit.
• In the other, the variable begins to execute
growing oscillations, where the amplitude is
increasing without limit.
• In both cases, some nonlinear breakdown (such
as an explosion or other malfunction) eventually
terminates the increase.
• A controlled variable in some process may be
stable and still cyclic.
• This is the case, for example, in two-position
control where the controlled variable oscillates
between two limits under nominal load conditions.
• A change in load may change the period of
oscillations , but the amplitude swing remains
essentially the same; hence, the variable is under
stable control.
Minimum Deviation : ( that is Mp % to be
minimised)
If a process-control loop has been adjusted to
regulate a variable at some setpoint value,
• then an obvious definition of quality is the extent to
which a disturbance causes a deviation from that
setpoint.
• Where a disturbance is a change in setpoint,
• this can be considered as any overshoot or
undershoot of the variable in achieving
• the new setpoint. In general, we want to minimize
any deviation of the dynamic variable from the
setpoint value. ( refer figure 9 a Underdamped case )
Minimum Duration : (ie “ ts ” to be minimised)
If a disturbance occurs, we can conclude that
some deviation will occur.
Definition of quality is the “ length of time
before the controlled variable regains the setpoint
value, or at least falls within the acceptable limits
of that value “.
( refer figure 9 a Under-damped case )
• Thus, the quality of a process-control loop is
defined through an evaluation of
stability, minimum deviation, and minimum duration
following a disturbance of the dynamic variable.
 Measure of Quality :
• In general, it is not enough simply to state that we
will design or adjust the process-control loop to
provide for stable, minimum-deviation, minimum-
duration operation.
Assume stable operation has been achieved.
There are three possible responses to a
disturbance that a dynamic variable in a process-
control loop can execute.
• The specific response depends on the controller
gains and lags in the process.
• Referring to Figure 9b for a load change and Figure
9a for a setpoint change, we have the following
definitions.
Overdamped :
• The loop is overdamped in case A of Figure 9; the
deviation approaches the setpoint value smoothly
(following a disturbance) with no oscillations.
• The duration is not a minimum in such a case. For
that matter, the deviation itself usually is not a
minimum either.
• Such a response is safe, however, in ensuring that
no instabilities occur and that certain maximum
deviations never occur.
Critically Damped :
Careful adjustment of the process-control loop
brings about curve B of Figure 9. In this case, the
duration is a minimum.
• This is the optimum response for a condition
where no overshoot or undershoot is desired
• in a setpoint change, or no cycling, in general, is
desired.
Underdamped :
The natural result of further adjustments of
the process-control
• loops is cyclic response, where the deviation
executes a number of oscillations about the
setpoint.
• This is shown in curve C of Figure 9. It is possible
that this response gives minimum deviation and
minimum duration in some cases. If the cycling
can be tolerated, then such a response is
preferred.
 Two specialized measures of control are
used when none of the above conditions serves to
define the measure of control desired in a process.
1 )Quarter Amplitude
2 ) Minimum Area
Quarter Amplitude criteria :
• When a process-control loop has a damped cyclic
response to a disturbance, a criterion is sometimes
used that is neither minimum deviation nor
minimum duration.
• Then quarter amplitude criteria is checked
• This measure of quality is found by adjusting the
loop until the deviation from a disturbance is such
that each deviation peak is down to one-quarter
of the preceding peak, as shown in Figure 10.
• In this case, the actual magnitude of the deviation
is not included in the measure, nor is the time
between each peak.
• In this sense, neither duration nor magnitude of
the deviation is directly involved in a quarter-
amplitude criterion.
Minimum Area :
• In cases of cyclic or underdamped response,
sometimes a combination of duration and
deviation, which must be minimized.
• Thus, if minimum deviation occurs at one loop
setting and minimum duration at another,then
neither is optimum.
• One type of optimum measure of quality in these
cases is to minimize the net area of the deviation
as a function of time.
• In Figure 11, this is shown as the sum of the
shaded areas. Analytically, this can be expressed
as
 STABILITY :
• Let us assumed that the loop response is stable
which is the prerequisite for use.
• In fact a great deal of effort is expended in the
design and development of process-control
loops to achieve this stability.
Transfer Function Frequency Dependence :
The static transfer function of an element
in a process-control loop tells how the output is
determined from the input when the input is
constant in time.
• The dynamic transfer function of an element tells
how the output is determined from the input
when the input varies in time.
• For the study of stability, we are interested in the
particular time variation that is sinusoidal
(i.e., the dynamic transfer function when the input is
oscillating at some frequency, f ).
Consider some element block as shown in Figure 13
with a transfer function “ T(ω)“, and where the
input is a sinusoidal given by
r = a sin(ω t)
• The frequency has been expressed in terms of the
angular frequency,2πf , measured in
radians/second.
• There are only two things that can happen, in a
linear study at least:
• The amplitude can change, and there can be a
phase shift. Thus, the output can be described as
c = b sin(ωt + Ø)
The ratio of the amplitudes is called the gain;
Gain = b / a
• and the phase shift is called the phase lag, Ø .
• In general, both the gain and amount of phase lag
of an element vary with frequency.
• The gain decreases, and the phase lag becomes
larger.
• The whole issue of stability is tied up with the
frequency variation of gain and phase of all
elements in a control loop.
Source of Instability :
• To see how a process-control loop can cause
instability, consider the open-loop block diagram
of Figure 14, for which an oscillating transient
• disturbance has been imposed at r.
• Notice that the feedback line has been broken at
the error detector,
• so no actual feedback occurs.
• Each element of the loop has a gain and phase
lag, including the process itself.
• The net gain is the product of all gains, and the
net phase lag is the sum of all phase lags.
• In a perfect world, the feedback, b, would be an
exact replica of the disturbance at
• every frequency. This would mean a loop gain of
unity and no phase shift. Then the -180
• phase shift (lag) of the error detector would
subtract the feedback from the disturbance, and
• it would be cancelled. In reality, there are gain
variations and extra phase shifts, and both
• vary with the disturbance frequency.
• If the gain for the disturbance frequency is
greater than one and the system phase lag small,
the disturbance is cancelled within a few cycles
of oscillation.
• However, as the phase lag becomes greater with
increasing frequency, the effectiveness of the
feedback is reduced, and the oscillation will
persist longer.
• Similarly, as the gain becomes smaller with
frequency, the effectiveness of the feedback
• to cancel error is reduced. But the control system
is still working and stable.
• Consider, however, the particular case of a
frequency where the phase lag of the system
• Reaches -180 while the gain remains unity or
greater. When combined with the errordetector
• phase shift, the net shift around the loop will be
360 , and so the feedback will be
• summed instead of subtracted. If the gain at that
frequency is just unity, then the disturbance will
persist forever, with constant amplitude.
• If the gain is greater than unity, the disturbance
will grow in amplitude. This is the instability
caused by the control system.
Instability Illustration:
• Instability is caused by a condition where for some
frequency the transfer function is such that
feedback to the error summer actually increases
the error because of the gain and phase shift.
• Now, if there is any frequency for which this
condition exists, then oscillations will always start
and grow at that frequency.
ILLUSTRATION :
• Assume that a small transient oscillation is
introduced from external sources as a disturbance
in r.
• Assume that at this frequency the gain is greater
than 1 ,say 2 and the phase shift is – 180 that is, a
lag of 180 . Let us study the result, frozen instant
by instant.
• The summer algebraically subtracts the feedback
from the input, giving the error signal e1.
• of the next instant in Figure 15a. In the next
instant, Figure 15b, the transient is gone but the
feedback,b2 , is the original amplified by 2 and
phase-shifted by 180 .
• This passes through the negative summing point
and becomes e2 . Figure15c shows e2
• amplified by 2 and phase-shifted by 180 , again
becoming b3 and then e3 . Thus, you can see
• that the error is actually growing! The control
system is forcing the oscillating error to increase,
• instant by instant. Of course, in actuality, it
happens smoothly, and the output results in
uncontrollable oscillations, for oscillating growth.
If there is any frequency where
• such conditions for growth exist, the system is
unstable, and something like random noise
• will eventually set the system into growing
oscillation.
• When a process-control installation is designed,
one has the objective of regulating the
controlled variable without instability
• in the loop. Stability can be ensured by designing
the controller gains so that oscillation growth is
never favourable, according to certain criteria
• Stability Criteria
• We can derive stability criteria by determining just
what conditions of system gain and phase lag can
lead to an enhancement of error (Unstable
system). We are led to two conclusions:
• 1. The gain must be greater than 1.
• 2. The phase shift must be 180 (lag).

• ( NOT TO BE USED )
• Thus, if there is any frequency for which the gain
is greater than one and the phase is -180 , the
system is unstable. From this argument,
• we develop two ways of specifying when a system
• is stable. These rules are as follows:
■ Rule 1 A system is stable if the phase lag is
more positive than at the frequency for which
the gain is unity (one).
■ Rule 2 A system is stable if the gain is less
than one (unity) at the frequency for which the
phase lag is 180 .
PROCESS LOOP TUNING : (setting optimum value
of Kp , Ki , Kd)

Two methods for tuning of the loop is described

which are as follows

1) Open Loop Transient Response Method ( for

self regulating process)(Process reaction
method)

2 ) Ziegler-Nichols Method ( for non self regulating

process)(Ultimate cycle method)
Open Loop Transient Response Method :
• This method of finding controller settings was
developed by Ziegler and Nichols and is sometimes
referred to as a process-reaction method.
• The basic approach is to open the process control
loop so that no control action (feedback) occurs.
This usually is done by disconnecting the controller
output from the final control element.
• All of the process parameters are held at their
nominal values.
• This method can be used only for systems with
self-regulation.
• At some time, a transient disturbance is introduced
by a small, manual change of the controlling
variable using the final control elements.
• The controlled variable is measured (recorded)
versus time at the instant of and following the
disturbance.
• A typical open-loop controller response is shown in
Figure 17, where the disturbance is applied at “t1 ”.
• Variations of controlled variable is obtained from
the graph.
Draw a tangent at inflection point .
The inflection point is defined as that point on the
curve where the slope stops increasing and begins to
decrease. Where the tangent line crosses the origin,
we get
L = lag time in minutes (7)
Figure17. from the graph we get T, the process
reaction time, and is given by
• Values of L and N used with the controlling
variable change “ΔP " to find the controller
settings. The following paragraphs give the stable
control
• definitions for the various modes as developed by
Ziegler and Nichols and corrections developed by
Cohen and Coon (when the quarter-amplitude
response criterion is indicated).
• In the latter case (Correction), a log ratio is used,
defined by
A transient disturbance test is run on a process loop.
The results of a 9% controlling variable change give a
process-reaction graph as shown in Figure 18. Find
settings for three mode action.
For the problem shown above find three mode
settings for quarter amplitude criterion