control performance
TRANSCRIPT
-
8/13/2019 Control Performance
1/14
8-1
CHAPTER 8
Control Performance 1
Once a control system has been designed and commissioned (started), it is necessary to monitor
it to assure continued performance. There are many reasons for changes in the performance of
control systems such as degradation in heat transfer areas due to fouling, mechanical problems
with valves and sensors, catalyst activations, and as we have mentioned several times in previous
chapters, nonlinearities. A couple of questions about installed working systems are, (a) how do I
know when my control system has problems? and (b) how can I achieve operational excellence
for my process control system?
Many metrics are available for this monitoring and for deciding if a control system is
performing to satisfaction. Sometimes we refer to them as Key Performing Index metrics, orKPI. We start with some of the traditional metrics and then move towards some of the most
recent ones in used now.
8-1 Traditional Evaluation Metrics
Chapters 3 to 5 mentioned that the great majority of open-loop systems respond to step changes
in forcing function as overdamped systems (monotonic, or non-oscillatory, response). However,
when a controller is installed and set in automatic (closed-loop) the response is commonly that of
an underdamped system (oscillatory response). For example, Fig. 8-1.1 may represent the
response of a closed-loop system to a unity step change in set point; Chapter 6 also showed some
of these types of response. Figure 8-1.2 shows a typical underdamped response. Chapter 2
presented and discussed this response, and for convenience of the reader we repeat the
presentation in the following paragraphs; the terms mentioned are metrics that have been
traditionally used in the performance evaluation of control systems.
-
8/13/2019 Control Performance
2/14
8-2
0 5 10 15 200
0.2
0.4
0.6
0.8
1
1.2
1.4
Time
Response,
y
.Underdamped
Overdamped
Critically damped
Figure 8-1.1 Possible responses to a set point Figure 8-1.2 Second-order underdampedchange response to step input
Per iod of Oscil lation; Decay Ratio; Rise Time; Settli ng Time; Overshoot
The specification of the desired value of the metric depends on the process and on what the
engineer thinks the control system should provide. A common value for the Decay Ratio is that
of and it is probably due to Ziegler Nichols tuning method (Chapter 6).
Another popular performance specification is the Integral Criteria; there are four of these
specifications. The Integral of the Absolute Error (IAE) provides an indication of the total erroras the integral is the area under the curve between the response and the set point, Fig. 8-1.3;
absolute value is used so that negative and positive dont cancel each other.
T
o
dtteIAE )(0 10 20 30 40 50 60 70 80
90
90.5
91
91.5
92
92.5
93
93.5
94
IAE (area under the
curve)
Time
Response
T
o
dtteISE2)(
T
o
dttetITAE )(
Figure 8-1.3 Integral of absolute value of the error (IAE) T
o
dttetITSE2)(
T
B
C
A
Rt
St
)(tY
Settling time
limits
t
-
8/13/2019 Control Performance
3/14
8-3
8-2 Key Metrics for an Installed Control System
The metrics presented in Section 8-1 have been traditionally used to evaluate the performance of
control systems. We now focus on the following five simple metrics and one troubleshooting
metric that we feel provide the necessary information. They are by no means inclusive but are
some of the ones we have used and feel confident.
% Time in Normal Mode - indicates if the loop is in the correct mode (auto, cascade) Average %ABS Error - indicates if the loop is controlling at set point Valve Travel, %/hrindicates if the valves are working too hard % Loops Oscillatingindicates if the loop is oscillating % Time Output is Saturatedindicates if there is an equipment sizing problem
Oscillation Periodthis is used as a troubleshooting aid
These metrics will advise/alert the operating personnel when work is required, what loops to
work on or who to call and when to call them. Knowing these metrics and acting to correct them
will stop wasting money and start saving money.
Next we discuss each metric and present the result of an audit that was performed at a
plant (for obvious reasons we do not divulge the name of the plant or any other information that
may be sensitive). An audit is a concentrated effort to analyze and tune loops at a site; fifty five
loops were considered. During the first visit, 26 loops were analyzed and tuned. During the
second visit, 29 loops were analyzed and tuned. In between visits, valves may have been worked
on that needed work, and control strategies may have been changed as recommended, with the
overall goal of improving the control system. The control system software used at the plant
provided an easy way to calculate the metrics.
During a normal operation (non-audit) the metrics are still active, and if a loop becomes
outside the specification, the control system will alert the operating personnel indicating which
loop has the problem.
8-2.1 % Time in Normal Mode The normal modeof a straight feedback controller is automatic, for the master controller of a
cascade system is automatic and for the slave controller is remote set (or cascade). These are the
modes required for complete benefit of automatic process control. If a controller is not in the
normal mode no action (or the wrong action) will be taken when a deviation of set point occurs.
-
8/13/2019 Control Performance
4/14
8-4
Figure 8-2.1 shows the % average time for all the loops in the plant including the result of
actions taken once the loops not in normal mode were identified.
0
10
20
30
40
50
60
70
80
90
100
Jan Feb Mar April May June July Aug Sept Oct Nov Dec
% Time in Normal Mode
26 loopsFirst visit 29 loops
Second visit
Figure 8-2.1 Average time in normal mode for all loops in the plant
At the start of the work, about 65% of the loops were in normal mode. After analyzing and
identifying the reasons the average rose to over 70% after the first visit and to about 95% after
the second visit.
8-2.2 Average %ABS ErrorFigure 8-2.2 shows the average % absolute error of all 55 loops; each error is calculated in %TO
(transmitter output).
-
8/13/2019 Control Performance
5/14
8-5
0
5
10
15
20
25
30
35
Jan Feb Mar April May June July Aug Sept Oct Nov Dec
Average % ABS Error
29 loops
Second visit
26 loops
First visit
Figure 8-2.2 % Absolute Error
At the start of the work the average % absolute error was about 30%, dropped to 25% after the
first visit, dropped further as work was continued after the first visit, and dropped to 12% after
the second visit. As work then progressed on repairing valves and implementing the
recommended changes to the control strategy, the end result was about 8%. Figure 8-2.3 shows
a sample of loops; it shows how the control system provides the information and the ease of
identifying the necessary loops requiring work.
30.96 30.6 25.11 18.32 17.17 11.95 8.972 7.118 6.389 6.652 7.722 7.756
Tagname Jan Fe b Mar April May June July Aug Sept Oct Nov Dec
FIC-2000 53.9 52.3 26.7 8.6 7.3 8.3 8.4 8.9 8.2 7.7 10.1 9.9
TIC-2303 5.6 5.7 5.4 4.8 4.7 4.5 3.2 3.1 2.7 2.7 3.2 3.0
LIC-2303 14.3 15.4 13.9 13.6 11.9 13.0 14.9 2.3 2.3 2.3 2.8 2.4
LIC-2402 16.7 15.4 11.9 5.8 4.6 5.6 5.3 6.0 5.2 5.8 6.0 6.1
TIC-3405 9.5 9.2 5.7 3.5 1.7 1.8 2.0 1.7 1.7 2.6 2.1 1.9
TIC-2404 6.2 6.2 5.8 5.3 4.9 5.1 1.9 1.9 1.8 1.7 2.0 2.1
FIC-2413 37.6 36.2 25.7 23.0 6.0 5.5 6.8 6.3 5.8 5.2 7.3 7.1
FIC-2613 53.9 49.5 27.6 17.0 9.0 9.8 10.3 10.4 8.9 10.2 10.7 10.9
TIC-3131 52.5 52.0 29.8 19.8 19.8 19.0 22.6 21.9 18.8 18.1 22.3 24.0
LIC-6201 30.8 30.5 30.3 29.0 28.9 12.0 5.2 3.9 4.1 4.8 5.3 5.1
TIC-12302 49.5 46.8 48.1 41.1 41.3 26.6 18.9 10.7 8.2 7.7 8.6 9.6
Average % ABS Error
Figure 8-2.3 Average % Absolute Error of individual loops
-
8/13/2019 Control Performance
6/14
8-6
8-2.3 Average Valve Travel
Figure 8-2.4 shows the progression in reduction of valve travel. At the start, the average travel
for the 55 loops was about 1150 %/hr. At the end of the work, this number was less than 200
%/hr, or a five-fold reduction.
0
200
400
600
800
1000
1200
1400
Jan Feb Mar April May June July Aug Sept Oct Nov Dec
Valve Travel %/Hr
Second visit
First visit
Figure 8-2.4 Average monthly valve travel
8-2.4 % Loops OscillatingFigure 8-2.5 shows the % of loops oscillating. In some cases oscillation may go up slightly after
tuning due to valve problems. The number goes down as control strategies are implemented and
valves repaired or replaced. It is necessary to point out that in some processes oscillations may
not disappear due to operating procedures such as start/stop machinery, purging of lines and
other equipment, etc.
-
8/13/2019 Control Performance
7/14
8-7
0
5
10
15
20
25
30
35
Jan Feb Mar April May June July Aug Sept Oct Nov Dec
% Oscillation
Second visit
First visit
Figure 8-2.5 % Loops Oscillating
8-2.5 % Time Controller Output is Saturated
In this context saturation refers to the controller output reaching one of its limits, either 0% or
100%. If the controller output is at its limit, or even including less than 5% or more than 95%,
the process is essentially out of control. Under a major or extreme disturbance the controller
may have to drive its output to one of its limits to compensate for the deviation. However, if the
% of time the output is saturated is significant (up to the operating people define what is
significant) it is an indication of a problem. Certainly the controller itself may have stopped
working but, if this is not the case, the one thing we can say is that the action taken by the
controller (increasing or decreasing its output) is not affecting the controlled variable and this is
why the controller keeps moving. The reason may be due to an upset operation, new operating
conditions that may require new equipment sizing, or a mechanical problem with the final
control element (common) or with the sensor, transmitter or the unit operation itself, etc.
8-2.6 Period of Oscillation
Figure 8-2.6 shows a number of loops that share a period of about 223 seconds. During the audit
the operating personnel claimed the analysis had to be incorrect because the loops were located
all through the plant and not anywhere near one another. But, they are all related to steam flow.
Thus there is a loop somewhere that is oscillating and these loops are seeing the oscillation as a
disturbance. In this case the period of oscillation provided a troubleshooting aid.
-
8/13/2019 Control Performance
8/14
8-8
TAG NAME DESCRIPTION
OSCILLATION
PERIOD SEC
21PC3775 50#F STEAM EXPORT 222.6
21PC2727 135 TO 50 LB STEAM 222.621PC2799 50 LB STM HEADER 222.6
21LC2517 DEAERATOR 222.7
21FC1399 STM TO X FRAC REBOILER 222.7
21FC2395 50 STM TO RERUN REBOILER 222.7
21FC1379 X4 REBOILER STEAM 222.7
21FC387 X3 REBOILER STEAM 222.7
21PC3720 X7 BURNER INJ 135# STEAM 222.8
21FC2185 STM TO GAS STABILIZER REBOILER 222.8
Figure 8-2.5 Period of Oscillation
8-3 Nonlinearities (revisited, once more)
In previous chapters we have presented and discussed the meaning and effect of process
nonlinearities. This is an important process characteristic and very much related to the topic of
this chapter; thus, a couple real examples are justified.
Example 8-3.1
Consider the heat exchanger shown in Fig. 8-3.1 where a process fluid is cooled using cooling
water. The process fluid flow varies often between 400 and 600 gpm. The temperaturecontroller was tuned at the 600 gpm level.
TV
TT
TIC
Cooling
Water
FI
Process
Fluid
Figure 8-3.1 Nonlinear process
-
8/13/2019 Control Performance
9/14
8-9
Figure 8-3.2 shows the controlled responses to a disturbance and to a set point change from 900F
to 92 0F and back to 90 0F. Both responses are stable and show small deviation from set point.Set point
changed from 90
to 92 F
0
10
20
30
40
50
60
70
80
50
60
70
80
90
100
110
0 500 1000 1500 2000 2500 3000 3500
ControllerOutputP
Va
nd
SetPoint
Time (sec)
Figure 8-3.2 Response to a disturbance and set point changes at 600 gpm
Figure 8-3.3 shows the responses when the process flow is changed to 400 gpm. This time
unstable responses are obtained. Obviously some of the KPI, such as absolute error or loop
oscillating, will alert operations.
0
10
20
30
40
50
60
70
80
50
60
70
80
90
100
110
0 500 1000 1500 2000 2500 3000 3500
ControllerOutput
PVa
ndS
etPoint
Time(sec)
Process flow changed
from 600 to 400 gpmUnstable response
Controller set inmanual
Set point
changed from 90
to 92 F
Figure 8-3.3 Responses at the 400 gpm level.
-
8/13/2019 Control Performance
10/14
8-10
The reason for the nonlinearity and therefore instability is that the process behaves differently at
different process flow rates. As the fluid flow slows through the cooler, the residence time
increases, so more heat is transferred; this makes the process gain larger. The time between when
the temperature valve moves and the measurement indicates a change is also longer; this is the
process dead time. Both gain and dead time affect the tuning as we have learned. Thus, this
situation requires re-tuning the controller
Example 8-3.2
Consider the reactor shown in Fig. 8-3.4 where an exothermic reaction occurs and it is cooled
using cooloing water. Depending on the type of product the reactor runs at temperatures
between Fo
119 and Fo149 . The most common temperature is F
o3.134 and the controller was
tuned for that condition.
TC
SP
Product
Reactor
Cooling
water
101
TR
TT101
Process
Fluid
Figure 8-3.4 Chemical reactor
Figure 8-3.5a shows the response of the temperature to a step change in set point of two
degrees up and down. The figure shows a stable fast response. Figure 8-3.5b shows the
response when the set point decreases by 10oF at a rate of 0.05
oF/min, and Figure 8-3.5c shows
a similar decrease by 13
o
F. Both responses are stable however, it seems that as the temperaturemoves away from the original set point of 134.3 oF where the controller was tuned, the process
response lags the set point more.
-
8/13/2019 Control Performance
11/14
8-11
0 50 100 150 200 250 300133
133.5
134
134.5
135
135.5
136
136.5
137
137.5
138
Temperature,
oF
Time, min
Reactor
Temperature
Set Point
Figure 8-3.5a Temperature response to set point changes
0 50 100 150 200 250 300 350 400 450 500 550124
126
128
130
132
134
136
Reactor
Temperature
Time, min
Tempera
ture,
oF
Figure 8-3.5b Process response to a decrease of 10oF in set point at a rate of 0.05
oF/min
-
8/13/2019 Control Performance
12/14
8-12
Set
Point
0 100 200 300 400 500 600 700 800120
122
124
126
128
130
132
134
136
ReactorTemperature
SetPoint
Time, min
Temperature,
oF
Figure 8-3.5c Process response to a decrease of 13 oF in set point at a rate of 0.05 oF/min
Figure 8-3.5d shows the response when the set point increases by 10oF at a rate of 0.05
oF/min, and Figure 8-3.5e shows a similar decrease by 13
oF. The first response, Fig. 8-3.5d,
shows a few oscillations at the new set point but recovers and controls. The response in Fig. 8-
3.5d is unstable. Obviously the process characteristics at this temperature are different enough
than before yielding an unstable system with the original tunings.
0 50 100 150 200 250 300 350 400 450 500 550
134
136
138
140
142
144
Time, min
Temperature,
oF
Reactor
Temperature
Figure 8-3.5d Process response to an increase of 10oF in set point at a rate of 0.05
oF/min
-
8/13/2019 Control Performance
13/14
8-13
0 50 100 150 200 250 300 350 400 450 500
134
136
138
140
142
144
146
148
150
Temperature,
oF ReactorTemperature
Time, min
Figure 8-3.5e Process response to an increase of 13oF in set point at a rate of 0.05
oF/min
-
8/13/2019 Control Performance
14/14