POWER June 2013 - 46

INSTRUMENTATION & CONTROL
element (for example, the control valve or
damper) is mechanically sticky. The combination
of a controller's integral action and a
sticky control element causes an oscillation
called a stick-slip cycle (Figure 2). When
loop oscillations are caused by stiction, the
controller output's trend often resembles a
saw-tooth or triangular wave, while the process
variable may look like a square wave or
an irregular sine wave.
Stiction can be detected by placing the
controller in manual mode and making small
changes (typically 0.5%) in controller output
followed by monitoring the process variable
for a resulting change. If the control valve
seems to accumulate a few of the controller
output changes before the process variable
responds to them all at once, the control
valve has stiction.
Deadband. Deadband in a final control
element is a mechanical defect causing a
dead zone through which the controller output
must change before the control element
responds after a change in travel direction.
Many control systems also provide an adjustable
deadband around a controller's set point.
The latter is sometimes used to prevent the
controller from reacting to process noise.
If an integrating control loop, such as mill
or feedwater heater level control, drives a
valve or damper directly, and the valve has
deadband, the loop will oscillate. An integrating
control loop will also oscillate if the
controller has a deadband feature that is set
wider than the measurement noise band.
Both kinds of deadband also reduce the effectiveness
with which a controller can eliminate
disturbances. Every time the direction of
a disturbance changes, the process variable
has to traverse the internal deadband before
the controller begins responding, and the
controller output has to traverse the controller's
final control element's deadband before
the latter actually begins moving. This causes
the control loop to perform sluggishly. The
speed of tuning should never be increased to
compensate for deadband, because the aggressive
tuning remains active after the dead3.
A deadband test. At time-slice A and B the controller output is at the same level. Because
of deadband, the process variable does not return to the same level. Source: OptiControls
Process variable
Controller output
band has been traversed, which can cause
stability problems. Deadband in final control
elements can severely affect the accuracy of
controller tuning, often resulting in overly
aggressive tuning.
Deadband can be detected with a simple
process test consisting of two controller
output steps in one direction and one step
in the opposite direction with the controller
in manual mode (Figure 3). The second and
last steps should be the same size to make it
easy to visually detect deadband from process
trends. If the process variable does not
reach the same level after the first and third
steps, it indicates the presence of deadband.
Final control elements with deadband must
be repaired for optimum control.
Tuning Control Loops
Incorrectly tuned control loops can respond
sluggishly, or they can overshoot and oscillate
significantly. Controller tuning is often done
in a trial-and-error way, but this is not nearly
as effective as using tuning rules or software.
Controller tuning should start with step tests
on the process to determine the dominant
process characteristics: process gain, dead
time, and time constant. Four or more step
tests should be done, the results compared,
outliers eliminated, and the average used for
tuning. If process characteristics change with
unit load or operating conditions, gain scheduling
should be used.
Note that using the IMC tuning method
Deadband
results in stable control loops, but it often
leaves a sluggish response to disturbances,
especially on slow loops like main steam
temperature. Cohen-Coon tuning provides
faster disturbance rejection, even with the detuning
recommended earlier. Quarter-amplitude-damping
tuning methods should not be
used because they result in marginally stable
loops and oscillatory plants.
When tuning control loops, it is important
4. Tune for response, not speed. A control loop becomes less stable and more oscillatory
with increasing speed of response. Source: OptiControls
Set point
Process variable
to understand that any control loop can hypothetically
be tuned for a slower response,
but that is not always the case for a faster
response. As a controller is tuned for a progressively
faster response, the control loop's
stability is compromised, and after some
point it begins oscillating (Figure 4). A control
loop's settling time is therefore limited to
a certain minimum. No control loop recovering
from a disturbance or responding to set
point change can settle out faster than this
minimum settling time.
The minimum settling time is virtually
Increasingly faster tuning
equal to the natural period of the process
being controlled. It will take a control loop
at least as long as the length of its process's
natural period to settle out after a disturbance
or set point change. The natural period of a
process is determined mostly by its effective
46
www.powermag.com
POWER | June 2013

http://www.powermag.com

POWER June 2013

Table of Contents for the Digital Edition of POWER June 2013