300 Process Control
300 Process Control
+
=
Fig. 300-12 Proportional-Plus-Integral Control Response (Open Loop)
Instrumentation and Control Manual 300 Process Control
Chevron Corporation 300-11 June 1989
Figure 300-13 depicts proportional-plus-integral control for a closed loop. In
response to a step change in load (top graph), the controlled variable (middle graph)
falls below the setpoint. The integral action adjusts the bias from 50% initially to
about 75% after the load change and shifts the position of proportional band
(shaded area) on the scale. Notice that the percentage value of the proportional
band is not changed. The lower graph shows the output of the controller.
Wind-up
A basic problem with integral controllers is that integral action continues as long as
an error exists. Assume a proportional-plus-integral controller is used to maintain
the level in the gas-liquid separator vessel in Figure 300-4. If a valve is closed
upstream of the vessel, the level drops below the setpoint. The controller then
closes the control valve in the outlet line to maintain the level setpoint. With no
inlet flow, the control valve closes completely and the vessel level is still less than
the setpoint.
A pneumatic control valve will typically be fully closed at a controller output of 15
psig. Since the measured vessel level is less than the setpoint, the integral action of
the controller continues to increase the controller output to the air supply pressure
(typically 20-30 psig). The action of the integral controller trying to exceed the
normal range of the controller output is called wind-up.
If the upstream valve is opened and flow is restored, the vessel level will rise above
the setpoint. The response of the controller to this high level will be delayed by the
wind-up. When the controller does respond, the output goes to the opposite limit. In
this case, the control valve will fully open and the vessel level will drop sharply.
The controller may oscillate through several cycles, stroking the control valve from
stop to stop on each cycle, before the oscillations cease and control is restored.
Fig. 300-13 Proportional-Plus-Integral Control Response (Closed Loop)
300 Process Control Instrumentation and Control Manual
June 1989 300-12 Chevron Corporation
Such oscillations overwork the control valve and, depending on the fluid and pres-
sures involved, can cause mechanical damage and seriously disrupt the process
downstream on the valve. An anti-wind-up feature may be included on controllers
that are frequently subject to this type of disturbance. This limits the controller
output range and thus prevents wind-up. When the process returns to normal, the
controller lag is eliminated and the oscillations are no worse than those in a propor-
tional controller.
Integral Time
Integral time should be proportional to the time it takes for the process to respond
to control action. When the process responds quickly, the integral time can be
shorter. If the integral time is too short, the control valve reaches its limit before the
measurement has time to respond. When the measurement does respond, it will
overshoot the setpoint, causing the integral to drive the valve to its opposite limit.
The time lag built into the gradual response of integral action lengthens the period
of oscillation of a loop. For a loop with proportional-plus-integral control, the
period of oscillation after a load change is longer than for proportional alone.
For loops where the exact value of the controlled variable is not critical, the shorter
period of the proportional-only controller can be an advantage. For example, a
vessel may operate within a wide range of liquid level without adversely affecting
pressure or gas quality. Therefore, the system level does not have to be accurately
controlled, and proportional control is often sufficient.
334 Derivative Control
With derivative action (also called rate action), the controller output is proportional
to the rate of change of the error. This means the faster the change in level, the
faster the change in controller output and control valve settings. By the same token,
if the level remains constant, even with a large error, the controller output would be
zero. This makes the use of derivative action by itself impractical.
Derivative action is normally combined with proportional action or proportional-
plus-integral action. Derivative action, being proportional to the rate of change of
the measured variable, introduces a lead (anticipation) element into the controller.
This increases the speed of response of the controller and compensates for the lags
introduced by proportional and integral actions. Figure 300-14 illustrates derivative
action.
The output from a proportional-plus-derivative controller may be expressed as
follows:
(Eq. 300-6)
where:
O = controller output
O K
c
E
n
T
D
M
n
M
n 1
S
-----------------------------
+ =
Instrumentation and Control Manual 300 Process Control
Chevron Corporation 300-13 June 1989
K
c
= controller gain
E
n
= error at time n
T
D
= derivative time, minutes
M
n
= measurement at time n
M
n-1
= measurement at previous sampling time
S = Time between measurements (sampling time)
The derivative action is greatest when integral and proportional action are just
beginning to respond. Derivative action also responds to the change in sign of the
measured variable. This opposes the tendency of integral and proportional action to
overshoot the setpoint and enables the controlled variable to settle out faster than
with either proportional or proportional-plus-integral action.
In Figure 300-14, area A represents the proportional component of controller
output. Note that the proportional response is a function of the difference between
the setpoint and the measured variable. Areas B and C represent the component
added or subtracted by derivative action. As the measured variable stops decreasing
and starts increasing, the sign of the derivative function changes. The integral
action (area D) eliminates offset by not returning to zero when the proportional and
derivative actions return to zero output. Areas E and F represent the corrections that
result from all three actions taken together.
Fig. 300-14 Proportional-Plus-Integral-Plus-Derivative-Control Action (Closed Loop)
300 Process Control Instrumentation and Control Manual
June 1989 300-14 Chevron Corporation
Derivative action, being sensitive to the rate of change of the measured variable,
cannot be used in processes that require fast response, or that have rapid fluctua-
tions or high noise levels. These conditions cause instability through large increases
in the derivative gain, and rapidly change direction (sign). Although derivative
action is difficult to tune because of its extreme sensitivity to measurement noise
and other high frequency disturbances, it does have some applications. Most impor-
tantly, it is used with proportional and integral action in temperature processes that
have large time lags.
Derivative action can be very helpful in controlling processes that have significant
deadtime, but using it can be difficult. Sometimes adding derivative action can
make the control loop appear slow and inactive with some types of process distur-
bances. This sluggishness might lead one to increase the amount of derivative and
perhaps also increase the controller gain. However, these new tunings might make
the controller unstable when a different disturbance occurs in the plant.
340 Advanced Control
Because this section of the Instrumentation and Control Manual is meant to be
introductory in nature, we will define the term advanced control to be anything
more sophisticated than simple, single-loop feedback control. Advanced control
would therefore include cascade control, feed forward control, signal selector
control, adaptive gain control, self-tuning controllers, multivariable control, matrix
control, and many other techniques too numerous to mention.
We will only deal here with cascade. The reader is encouraged to consult the refer-
ences listed in Section 360 for additional information. The Monitoring and Control
Systems Division in the Engineering Technology Department is also available for
consultation.
341 Cascade Control
Cascade control should also be considered when the primary control variable is
slow to react to disturbances. Like any feedback control loop, a cascade control
loop has a controlled variable, a setpoint and a controller. However, instead of
having a valve as its final control element, a cascade controller sends its output to
the setpoint of another controller, adjusting this setpoint to correct an error in the
controlled variable. This other controller is called the secondary or slave
controller. The cascade controller is called the primary or master controller.
If disturbances in the process can be recognized and quickly corrected, the primary
control loop will not be affected. This suggests that the secondary control loop must
operate faster than the primary loop. In fact, general guidelines suggest that the
secondary loop should respond at least five times faster than the primary loop.
Looking again at the example of the furnace, let us assume that the fuel system
provides fuel to several other furnaces as well. Over the course of hours, the pres-
sure in the system might well vary as the fuel demand in all of the furnaces
changes. A change in the fuel header pressure changes heat transfer in the furnace.
Instrumentation and Control Manual 300 Process Control
Chevron Corporation 300-15 June 1989
Because heat transfer is a slow process, the outlet temperature controller cannot be
tuned well enough to eliminate the effect of changing fuel flow (see Figure 300-15).
(For details on controller tuning, see Section 350.)
On the other hand, if the fuel flow remains steady while the pressure is changing,
the furnace temperature will be more constant. Fuel flow changes almost immedi-
ately when the control valve is moved. Therefore, the flow controller can be tuned
to eliminate most of the disturbances in fuel flow.
Such circumstances lend themselves to the use of cascade control: a fast process
(fuel flow), a slow process (furnace heat transfer), and a disturbance (fuel pressure)
that affects the fast loop. Figure 300-16 shows the cascade control system for the
furnace.
Fig. 300-15 Feedback Control Performance
Fig. 300-16 Cascade Control
300 Process Control Instrumentation and Control Manual
June 1989 300-16 Chevron Corporation
Compare Figure 300-17 to 300-15. With cascade control the outlet temperature is
much more steady. The fuel gas controller (secondary controller) has eliminated
almost all fuel pressure disturbance from the furnace.
342 Feed-forward Control
Feed-forward control measures a disturbance before it can affect the controlled vari-
able, and changes the manipulated variable to compensate for the disturbance. Of
course, for feed-forward control to work properly, the magnitude and timing of the
effect on the controlled variable must be known. The process might be worse off if
the manipulated variable is changed too much or too quickly.
In Figure 300-18, a gas-fired furnace process is equipped with a temperature
controller (TC), a feed-forward controller (FFC), and a summer, which adds the
two controller outputs together. The feed-forward controller, also called a flow frac-
tion controller, operates like a simple multiplier: The output of the FFC consists of
its input (from the flow transmitter FT) multiplied by a ratio entered by the operator.
Fig. 300-17 Cascade Control Performance
Fig. 300-18 Feed-forward Control
Instrumentation and Control Manual 300 Process Control
Chevron Corporation 300-17 June 1989
Figure 300-19 shows what might happen in a real furnace as the feed rate is
changed. In the top graph, the feed rate to the furnace is raised at time 1. By time 2,
the furnace outlet temperature begins to drop below setpoint. The fuel valve then
begins to open and raises the outlet temperature back to the setpoint by time 3. In
the bottom graph, the fuel valve has begun to open by time B, and by time C the
furnace temperature is back to the original setpoint. With feed-forward and feed-
back control, the process has recovered from the feed rate disturbance much faster
than with feedback control alone. Note that the temperatures period of oscillation
is the same in both cases. This period is a dynamic characteristic of the furnace and
cannot be changed by the control system. However, the feed-forward controller has
been able to reduce the size of the temperature disturbance and has speeded up the
recovery.
Feed-forward control should not be used by itself, but always with feedback
control, because the rate and magnitude of the reaction of a process to a disturbance
is rarely consistent.
Fig. 300-19 Feedback/Feed-forward Control Performance
300 Process Control Instrumentation and Control Manual
June 1989 300-18 Chevron Corporation
350 Controller Tuning
Several methods are available to tune a controller to function in a specific loop. The
following discussion considers some of the methods commonly used. Several of the
references in Section 360, particularly Reference 5, should be useful when difficult
situations are encountered.
351 Quarter Decay Method
The quarter decay method is a closed loop controller tuning method. This means
that the controller remains in automatic while tuning adjustments are made.
The quarter decay method defines the ultimate limit for tight controller tuning.
Often, the tuning constants it produces are too tight (too sensitive) in processes that
have sticky valves and noisy measurements.
To prevent controllers from going unstable unexpectedly, tuning constants should
be set to values one-half as sensitive as those obtained with the quarter decay
method. After these less sensitive tunings are exposed to actual upsets and irregular-
ities, and the operators gain confidence in the controller tuning, it may be appro-
priate to make the tunings more sensitive.
The general tuning sequence is as follows:
1. With the controller in automatic, adjust all tuning constants to their least sensi-
tive (least effective) setting. Proportional band should be at its highest value
(proportional gain should be at its lowest value). Integral time should be at its
highest value (most minutes per repeat or least repeats per minute). Derivative
time should be at its highest value.
2. Make a small step change in controller setpoint and record the controller
measurement until it settles out.
3. Change the setpoint back to its original value. Record the measurement as
before.
4. Increase the proportional gain (reduce the proportional band) in small steps
and repeat steps 1-3 until the recording of the output resembles Figure 300-20,
curve B; that is, until the amplitude of the first positive excursion of curve B is
approximately four times that of the second (thus the name, quarter decay
method).
5. Measure the period of oscillation. Set the reset and derivative:
T
R
= P/1.5 minutes
(Eq. 300-7)
T
D
= P/6 minutes
(Eq. 300-8)
Instrumentation and Control Manual 300 Process Control
Chevron Corporation 300-19 June 1989
6. With T
R
and T
D
set at above values, reestablish controller gain for quarter
decay.
Figures 300-21, 300-22, and 300-23 show how the three tuning parameters affect
the response of a controller. With proportional-only control, settling time is fairly
long and there is a permanent offset from the setpoint. Adding integral control
reduces settling time and eliminates offset. Adding derivative control to propor-
tional control reduces settling time but not offset. Only integral control eliminates
the offset.
Fig. 300-20 Quarter Decay Method Tuning
Fig. 300-21 Proportional-only Controller Response Fig. 300-22 Proportional-Plus-Integral Controller
Response
300 Process Control Instrumentation and Control Manual
June 1989 300-20 Chevron Corporation
352 Ultimate Sensitivity Method
The ultimate sensitivity method ( Figure 300-24) is also a closed loop test. Adjust
the integral time and/or the derivative time to their minimum values. Then narrow
the proportional band (increase gain) in small steps, each time changing the
setpoint as described in Section 351, until the controller measurement just begins to
cycle continuously. This proportional band setting is called the ultimate propor-
tional band, denoted PB
u
. The period of oscillation at the ultimate proportional
band is called the ultimate period, measured in minutes and denoted P
u
. The
amplitude of the oscillations in Figure 300-24 has been exaggerated for clarity.
The ultimate proportional band, PB
u
, and the ultimate period, P
u
, are then used to
calculate tuning constants as shown in Figure 300-25. These constants give the
quarter damping response already discussed.
Note that Figures 300-25 and 300-26 show two sets of equations for a proportional-
plus-integral-plus-derivative controller. The set identified as Commercial should
be used for controllers encountered in industry. The set identified as Ideal is
based on an ideal control algorithm equation commonly used in universities. They
are included here for completeness.
353 Process Reaction Curve Method
This is an open loop tuning method. The controller remains in manual while
response tests are made. The tuning method measures two parameters to describe
the response characteristic of the process: process deadtime and process time
constant.
The deadtime is the delay between a change in valve position and the resulting
change in the controlled variable. The process time constant is the time required for
the controlled variable to reach approximately 60% of its final value.
Fig. 300-23 Proportional-Plus-Derivative Controller Response
Instrumentation and Control Manual 300 Process Control
Chevron Corporation 300-21 June 1989
To perform this test, change the controller valve position by a small amount and
record the controlled variable. The deadtime, TD, and time constant, TC, are
measured and their values used to calculate the controller tuning constants.
Figure 300-26 shows how the measurements are made and used.
Fig. 300-24 Ultimate Sensitivity Method
Fig. 300-25 Ultimate Sensitivity Method Tuning Constants
Proportional Band
(%)
Reset Time
(minutes)
Derivative Time
(minutes)
Proportional Controller 0.5 PB
u
Proportional + Integral Controller 0.45 PB
u
P
u
/ 1.2
Proportional + Integral + Derivative
Controller
Ideal
0.6 PB
u
P
u
/ 2.0 P
u
/ 8.0
Proportional + Integral + Derivative
Controller
Commercial
0.3 PB
u
P
u
/ 4.0 P
u
/ 4.0
Notes PB
u
= Ultimate Proportional Band, %
P
u
= Ultimate Period, minutes
300 Process Control Instrumentation and Control Manual
June 1989 300-22 Chevron Corporation
Note that the process reaction curve method cannot be used to integrate processes
such as level control; when a valve controlling a level is changed the level
continues to change until the vessel overflows or empties. Level controllers can be
tuned using the ultimate sensitivity method or more advanced methods discussed in
Reference 5.
Figure 300-27 gives typical ranges of controller tuning constants for various
processes. Use these values with caution; your process might not be typical. The
Fig. 300-26 Open Loop Reaction Curve Method Tuning Constants
Instrumentation and Control Manual 300 Process Control
Chevron Corporation 300-23 June 1989
exact values must be determined by one of the above methods. For future reference,
always record the control loop ID number (e.g., FRC-123), the date, and the tuning
constant when you have finished tuning a control loop.
360 References
1. Fundamentals of Process Control Theory. Instrument Society of America,
1981.
2. Process Control Systems. McGraw-Hill, 1979.
3. Process Instruments and Controls Handbook. McGraw-Hill, 1974.
4. Controllers & Control Theory. Production Facility Bookware Series, Interna-
tional Human Resources Development Corp., 1987.
5. Tuning and Control Loop Performance. Instrument Society of America, 1983.
Fig. 300-27 Tuning Constants for Typical Process
LOOP TYPE
PROPORTIONAL
BAND %
RESET TIME
(MINUTES)
DERIVATIVE TIME
(MINUTES)
Flow 100 - 500 0.02 - 0.1 none
Liquid Pressure 100 - 500 0.02 - 0.1 none
Gas Pressure 1- 50 0.1 - 0.5 none
Level 1- 50 0.05 - 0.25 none
Temperature 10 - 100 1 - 10 0.5 - 20