You are on page 1of 28

Chemical Engineering Process Laboratory III

Feed Forward and Cascade Control Experiment

Summary
In this experiment, the effects of the feedforward and cascade control were
investigated. The CE2000 Control software was used to set the parameters of the two
controllers and to monitor the process vessel level while the experiment was run on a
CE117 Process Trainer.
In the first part of the experiment, the effects of process disturbance on feedforward
control and its limitation were studied. The system was set up for proportional-

integral (PI) control of the process vessel using the valve position as the controlled
variable. Once the system had stabilized, the pump voltage was changed to introduce
a disturbance to the process. The action of the feedforward control on the process as it
corrected the disturbance was noted. This was repeated several times with
feedforward gains of 0.2, 0.4, 0.6 and 0.8 respectively. From the results, it was found
that a feedforward control with a gain of 0.2 exhibited the most optimum
performance.
In the second part of the experiment, the actions of the cascade control on the process
system when a disturbance is introduced, were studied. The system was again set up
for PI control for both the master and slave controllers. Different types of disturbances
were introduced separately throughout the experiment which include disturbances to
the set point, changes to inflow rate and changes to outflow rate. The effects of the
cascade control for each of the disturbances were noted.

Contents

1. Introduction................................................................................................................1
2. Theoretical Background.............................................................................................2
2.1 Feedforward Control............................................................................................2
2.2 Cascade Control....................................................................................................3
3. Experimental Setup and Procedures...........................................................................4
3.1 Experimental Setup..............................................................................................4
3.2 Experimental Procedures......................................................................................5
3.2.1 Part 1: Feedforward Control..........................................................................5
3.2.2 Part 2: Cascade Control.................................................................................6
4. Results and Discussion...............................................................................................8

4.1 Feedforward Control Experiment.........................................................................8


4.1.1 Feedforward Gain = 0....................................................................................8
4.1.2 Feedforward Gain = 0.2.................................................................................9
4.1.3 Feedforward Gain = 0.4...............................................................................10
4.1.4 Feedforward Gain = 0.6...............................................................................11
4.1.5 Feedforward Gain = 0.8...............................................................................12
4.2 Cascade Control Experiment..............................................................................14
4.2.1 Disturbances to Set Point Change................................................................14
4.2.2 Disturbances to Inflow Rate........................................................................16
4.2.3 Disturbances to Outflow Rate......................................................................19
4.3 Discussion...........................................................................................................21
4.3.1 Feedforward Control....................................................................................21
4.3.2 Cascade Control...........................................................................................22
5. Error Analysis...........................................................................................................23
6. Lab Safety Analysis..................................................................................................24
7. Conclusion................................................................................................................24
8. References................................................................................................................25

1. Introduction

Process control is a key component in operating an industrial process plant. In a


typical plant, many complicated processes could be running at the same time with an
even larger number of process variables that must be measured and controlled. It is
extremely difficult to continuously monitor and prevent disturbances to the processes
manually. A computer-based process control system is thus needed to operate the plant
safely while ensuring the quality of the product. The role of process control is to
therefore maintain the process variable near the desired values and to automatically
correct disturbances introduced to the processes.
One classic and simple process control is the feedback control. In feedback control,
the controlled variable is measured and this is used to adjust the manipulated variable.
It is thus a reactive control system as it requires information on deviation from set
point to take action.
Although a feedback control is simple to implement and requires minimum
information to operate, it has several inherent disadvantages. Firstly, the feedback
control is unable to prevent deviation of the controlled variable from the set point
during a disturbance. This is because it can only take corrective action only after a
disturbance has occurred. Secondly, the feedback control may not be suitable for
processes with large time constants and/or long time delays. Such processes may not
be able to achieve steady-state if large disturbances occur frequently.
With these important drawbacks of the feedback control, other types of process
control strategy would need to be looked into to overcome these disadvantages. Two
process controls, feedforward control and cascade control, both address the
shortcomings of the feedback control. These two process control employ different
strategies to correct disturbances while overcoming the disadvantages of the feedback
control.
For this study, the feedforward control and the cascade control were investigated to
understand their actions on the processes when disturbances are introduced. The
objectives of the experiment are to study effects of process disturbance on
feedforward control and its limitation and to study how cascade control can improve
the control.

2. Theoretical Background
2.1
Feedforward
Control
In feedforward control, the disturbance variable is measured instead of the controlled
variable. This allows the feedforward control to take corrective action, based on its
measurements of the disturbance variable, before the disturbance causes any deviation
from the set point in the controlled variable. This would ideally mean that the
feedforward control makes adjustments to the manipulated variable such that it
cancels the effect of the disturbance to ensure that the controlled variable remains
unaffected. The feedforward control is thus a proactive control system which is in
contrast with the feedback control as it is a reactive control system. A schematic
diagram of a feedforward control is as illustrated in Figure 1.

Figure 1. Schematic diagram of feedforward control. The feedfoward controller measures the
disturbance variable and make adjustments in the manipulated variable to correct the disturbance
before deviations from the set point occur

Despite being able to overcome the feedback control main drawback, the feedforward
control has several disadvantages as well. One of them is that the disturbance variable
must be measured on-line which is not feasible in many situations. Another
disadvantage is that it needs to have information on the process model in order to
work effectively and its performance is thus dependent on the accuracy of the process
model. The process model may not be known all the time and at times, only an
approximate process model can be obtained. The feedforward control would thus be
unable to correct for disturbances that are not accounted for and could only correct for
the disturbances that it is measuring.
These disadvantages led to the use of feedforward control in combination with
feedback control in practical applications. Feedforward control would measure the
disturbance variables and make corrective actions to decrease the effect of these
measured disturbances while the feedback control help to correct for disturbances that
are not measured by the feedforward control and compensates for the inaccuracies in
the process model. A typical feedforward-feedback setup is shown in Figure 2.

Figure 2. Block diagram of a feedforward-feedback control system. The feedback trim is added to
compensate for inaccuracies in the process model and unmeasured disturbances.

2.2 Cascade Control


The cascade control is another process control strategy which also addresses the
shortcoming of the feedback control. In a cascade control system, two feedback
controllers are used. One acts as the master controller which measures the primary
controlled variable and uses the primary measurements to establish the set point for
the secondary controller, known as the slave controller. The slave controller measures
the secondary variable and the output of the slave controller adjusts the manipulated
variable. It is located in the secondary loop that is nested within the primary loop.
This is illustrated in Figure 3 below where the master controller in the primary loop
measures the temperature of the liquid in the reactor and determines the set point for
the slave controller in the secondary loop which is measuring the temperature of the
jacket. The slave controller then adjusts the cooling water valve accordingly.

Figure 3. Cascade control system implemented for a chemical reactor. The primary measurement is the
temperature of the chemical in the reactor while the secondary measurement is the temperature of the
cooling jacket.

The main advantage of the cascade control is that the second measured variable is
located close to a potential disturbance. This allows the secondary controller to detect
the disturbance sooner than the primary controller and corrects it immediately. The
cascade control thus responds to disturbances faster than a feedback control. This also
makes the cascade control less prone to errors in the process model.
3. Experimental Setup and Procedures
3.1 Experimental Setup
The experimental setup includes a CE117 Process Trainer (shown in figure 4), a CE117
Mimic Panel (shown as figure 5) and the CE2000 software.

Air vent

Water tank

Water reservoir

Process drain valve

Pump
Figure 4. CE117 Process Trainer

Level Transmitter

Flow Transmitter

Proportional Valve
Process vessel

Pump

Figure 5. CE117 Mimic Panel

3.2 Experimental Procedures


3.2.1 Part 1: Feedforward Control
Connections for the mimic panel are the same as lab manual (shown in Figure 6).

Figure 6. Connections for Feedforward Control

Procedures for feedforward control process:


Starting CE2000 software and loading the file C3 Feedforward Control.ict.

me as figure 6; set heater control to Manual and at minimum power; set pump 2 to external; turn off cooler fan a

Set feedforward gain to 0, and PID controller is set to 5, 0.2 and 0 respectively.
Set level set point and pump voltage set to 7V.

Start software by pressing the record button. System is allowed to settle within 10% of the level set point

Change pump voltage from 7V to 10V, and wait the system to reach steady state.
Change pump voltage from 10V to 7V, and wait the system to reach steady state.
Press the stop button and export the data.

The process is repeated for another 4 times by changing only the feedforward gain to 0.2, 0.4, 0.6 and 0.8

3.2.2 Part 2: Cascade Control


Similarly, follow the lab manual to do connections for the mimic panel (shown as Figure 7).

Figure 7. Connections for Cascade Control

Start CE2000 software and load the file C3 Cascade Control.ict.


Close process loop bypass valve, and both air vent and process vessel drain are opened fully.

Set voltage to the proportional valve to 10V and the reference level to 7V.
Master PID controller settings are set to 20, 1 and 0 respectively.
Slave PID controller settings are set to 1, 1 and 0 respectively.

Open process vessel drain valve; set pump 2 to external and run the software by pressing record button

level set point from 7VChange


to 8V, and
voltage
wait to
system
the proportional
to reach steady
1/3
valve
Close
state.
from
the
10V
process
to 7.5,
drain
waitvalve
for stabilized.
manually and wait for

Fully
open
the
valve and
for steady
Stop
system
n the system isChange
stabilized,
voltage
change
from
the
7.5V
level
to set
10Vpoint
and
wait
from
for
8V stabilized.
to
7V.
Stopwait
the system
andstate.
export
thethe
data.

When it reaches steady state, stop it and


Repeat
export
by changing
the data. voltage
Repeat
to 5V
by changing
and 2.5V. the valve to 1/2 closed and 2/

4. Results and Discussion


4.1 Feedforward Control Experiment
4.1.1 Feedforward Gain = 0

FF Controller Gain = 0
40
30

Setpoint
Control

Voltage (V) 20

Level

10
0
0

100 200 300 400 500 600 700 800


Time (s)

Description
Period

Initial Steady State Phase

Disturbance: 7V to 10V

Disturbance: 10V to 7V

0.0 368.1s
368.2 488.6s
488.6 717.5s
Figure 8. Plot of Responses when Feedforward Gain = 0

There is no feedforward controller action in this system. In the experiment, the tank was
initially filled from 0 to the desired setpoint gradually with an overshoot to at 59.4s. The
action of PI controller then slowly brought the system level back to the setpoint slowly during
59.4s to 123.90s.
At 368.2s, a step increase of disturbance from 7V to 10V was introduced and it resulted in an
offset of 0.12V which lasted until 346.2s before the PI controller takes action and eliminates
the offset. During this period, the control voltage decreased gradually to 2.45V and increased
back to 2.88V and oscillated around that steady state.
At 488.6s, a step decrease of disturbance from 10V to 7V was introduced and it resulted in an
offset of -0.12V which lasted until 506.4s before the PI controller takes action and eliminates
the offset. During this period, the control voltage increased gradually to 3.9V and decreased
back to 3.6V and oscillated around that steady state.

Table 1: Summary of Feedforward Control with 0 Gain

Feedforward Gain = 0
Setpoint (V)

Level (V)

Controller (V)

Initial Steady

7.097

3.441

7.000

Time to reach
steady state (s)
123.90
9

State
7V 10V
10V 7V

7.000
7.000

+/- 0.002
+/- 0.002

2.788
3.507

54.60
111.00

4.1.2 Feedforward Gain = 0.2

FF Controller Gain = 0.2


40
35
30
Voltage (V)

25

Setpoint

20

Control

15

Level

10
5
0
0

100

200

300

400

500

600

700

Time (s)

Description
Period

Initial Steady State Phase

Disturbance: 7V to 10V

Disturbance: 10V to 7V

0.0 374.8s
374.8 507.7s
507.8 589.1s
Figure 9. Plot of Responses when Feedforward Gain = 0.2

In this system, there is a P controller acting as a feedforward controller with a gain of 0.2. The
experiment was repeated. At the initial phase, there was a slight overshoot to at 58.1s before
the system level was brought back to the setpoint during 58.1s to 110.1s.
At 374.8s, a step increase of disturbance from 7V to 10V was introduced and it resulted in a
negligible offset of less than 0.02V from the setpoint. During this period, there was a sharp
drop in control voltage from 3.55V to 2.8 V. This is because the feedback controller paired
with the feedforward control action is able to eliminate the offset. The level value was then
brought back to the setpoint value through small and gradual oscillations.
At 507.8s, a step decrease of disturbance from 10V to 7V was introduced. Once again, there
was a very negligible offset of less than -0.02V from the set point due to the pairing of the
feedback and feedforward controller. There was a sharp increase of control voltage from 2.8V
to 3.6V. The level value was then brought back to the setpoint value through small and
gradual oscillations.

Table 2: Summary of Feedforward Control with 0.2 Gain

Feedforward Gain = 0.2


10

Initial Steady
State
7V 10V
10V 7V

Setpoint (V)

Level (V)

Controller (V)

7.000

7.098

3.435

Time to reach
steady state (s)
110.10

7.000
7.000

+/- 0.002
+/- 0.002

2.794
3.612

negligible
negligible

4.1.3 Feedforward Gain = 0.4

FF Controller Gain = 0.4


40
30
Setpoint
Voltage (V)

Control

20

Level
10
0
0

100

200

300

400

500

600

700

Time (s)
Description
Period

Initial Steady State Phase

Disturbance: 7V to 10V

Disturbance: 10V to 7V

0.0 268.6s
268.7 448.6s
448.6 648.5s
Figure 10. Plot of Responses when Feedforward Gain = 0.4

The feedforward controller gain was increased from 0.2 to 0.4 as compared to the previous set
up. At the initial phase, there was an overshoot to at 67.0s before the level was brought back
to the setpoint during 67.0s to 122.5s. It can also be observed that there is a larger oscillation
of the control voltage and a longer time is needed to attain steady state as compared to the 0.2
gain.
At 268.7s, a step increase of the disturbance from 7V to 10 V resulted in an offset of less than
-0.05V from the setpoint. During this period, there is a sudden drop in the control voltage
from 4.4V to 3.2V. At 714.3s, a step decrease of the disturbance from 10V to 7V resulted in
an offset of less than 0.03V from the set point. During this period, there is a sudden rise in the
control voltage from 3.46V to 4.66V. In both cases, the offset is eliminated by the pairing of
the feedforward controller action to the feedback controller. The controller then brought the
level back to the setpoint of 7V via small and gradual oscillations.

Table 3: Summary of Feedforward Control with 0.4 Gain

Feedforward Gain = 0.4


Setpoint (V)

Level (V)

Controller (V)

Time to reach
steady state (s)
11

Initial Steady
State
7V 10V
10V 7V

7.000

7.099

4.261

122.50

7.000
7.000

+/- 0.002
+/- 0.002

3.556
4.277

63.60
39.80

4.1.4 Feedforward Gain = 0.6

FF Controller Gain = 0.6


35
30
25

Setpoint

20
Voltage (V)

Control

15

Level

10
5
0
0

100

200

300

400

500

600

Time (s)
Description
Period

Initial Steady State Phase

Disturbance: 7V to 10V

Disturbance: 10V to 7V

0.0 234.2s
234.2 370.3s
370.3 488.3s
Figure 11. Plot of Responses when Feedforward Gain = 0.6

The feedforward controller gain was increased from 0.4 to 0.6 as compared to the previous set
up. At the initial phase, there was an overshoot at 63.1s before the level was brought back to
the setpoint during 63.1s to 123.5s. It can also be observed that there is a larger oscillation of
the control voltage and a longer time is needed to attain steady state as compared to the 0.2
gain.
At 234.2s, a step increase of the disturbance from 7V to 10 V resulted in an offset of less than
-0.15V from the setpoint. During this period, there is a sudden drop in the control voltage
from 4.44V to 2.66V. At 370..s, a step decrease of the disturbance from 10V to 7V resulted in
an offset of less than 0.15V from the set point. During this period, there is a sudden rise in the
control voltage from 3.55V to 5.35V. In both cases, the offset is eliminated by the pairing of
the feedforward controller action to the feedback controller. The controller then brought the
level back to the setpoint of 7V via small and gradual oscillations.

Table 4: Summary of Feedforward Control with 0.6 Gain

Feedforward Gain = 0.6


Setpoint (V)

Level (V)

Controller (V)

Time to reach
12

Initial Steady
State
7V 10V
10V 7V

7.000

7.099

4.312

steady state (s)


123.50

7.000
7.000

+/- 0.002
+/- 0.002

3.552
4.341

45.00
58.10

4.1.5 Feedforward Gain = 0.8

FF Controller Gain = 0.8


35
30
25

Setpoint

20
Voltage (V)

Control

15

Level

10
5
0
0

100

200

300

400

500

600

Time (s)

Description
Period

Initial Steady State Phase

Disturbance: 7V to 10V

Disturbance: 10V to 7V

0.0 274.3s
274.4 374.4s
374.5 524.8s
Figure 4. Plot of Responses when Feedforward Gain = 0.8

The feedforward controller gain was increased from 0.6 to 0.8 as compared to the previous set
up. At the initial phase, there was an overshoot at 63.3s before the level was brought back to
the setpoint during 63.3s to 124.6s. It can also be observed that there is a larger oscillation of
the control voltage and a longer time is needed to attain steady state as compared to the 0.2
gain.
At 274.3s, a step increase of the disturbance from 7V to 10 V resulted in an offset of less than
-0.25V from the setpoint. During this period, there is a sudden drop in the control voltage
from 4.4V to 2.0V. At 374.5s, a step decrease of the disturbance from 10V to 7V resulted in
an offset of less than 0.24V from the set point. During this period, there is a sudden rise in the
control voltage from 3.6V to 6.0V. In both cases, the offset is eliminated by the pairing of the
feedforward controller action to the feedback controller. The controller then brought the level
back to the setpoint of 7V via small and gradual oscillations.

Table 5: Summary of Feedforward Control with 0.8 Gain

Feedforward Gain = 0.8


Setpoint (V)

Level (V)

Controller (V)

Time to reach
steady state (s)
13

Initial Steady
State
7V 10V
10V 7V

7.000

7.097

4.369

124.60

7.000
7.000

+/- 0.002
+/- 0.002

3.595
4.383

58.40
86.60

From the data collected, it can be observed and concluded that a feedforward controller gain
of 0.2 is the most optimum for the system.
A large oscillation in the control voltage and huge offset in the level voltage are the outcomes
of the corrective action taken by the feedback controller without the feedforward controller,
i.e. gain = 0. In addition, a large amount of time is needed for the level voltage to approach
back to setpoint. With the implementation of 0.2 feedforward controller gain, the corrective
action becomes gentler and resulted in a negligible offset in the level voltage. Also, the
control voltage experienced a smaller and more gradual oscillation. However, when the
feedforward controller gain is increased beyond 0.2, i.e. 0.4, 0.6 and 0.8, the corrective action
becomes significantly more vigorous and there is a larger offset in the level voltage. The time
taken for the level to approach back to setpoint after the disturbance also increased when the
feedforward controller gain increased.
Interests were put in to investigate the controller behaviour of different feedforward controller
conditions. When there is a disturbance introduced, the controller voltage increased or
decreased sharply in the opposite direction before corrective action taken to stabilize the
system. For example, when there is a step increase in voltage (disturbance), the controller
voltage decreased sharply, and vice versa. Comparing all 5 conditions, i.e. when feedforward
controller gain is 0, 0.2, 0.4, 0.6 and 0.8, it showed the system with when the feedforward
controller gain (= 0.2) will bring a smallest change in control voltage when disturbance was
introduced. This could also explain why the level offset for 0.2 feedforward controller gain is
negligible.
Hence, it could be concluded that enhancement in the performance of the system is ensured
for a well-tuned feedforward and feedback controllers pairing system. In this case, the
optimized feedforward controller gain for this system is around 0.2. To obtain an exact value
for the optimized feedforward controller gain, further fine tuning and more experiment could
be
conducted.

4.2 Cascade Control Experiment


While feedforward control is able to compensate for the disadvantages of using conventional
feedback control, especially for processes with large time delays or time constants, using
feedforward control requires explicit knowledge and quantification of the disturbances, which
is impossible in several real life cases. In this section, we will investigate the effectiveness of
using cascade control as an alternative to feedforward control.
Cascade control is a control configuration that consists of a secondary controller, also known
as a slave controller, nested within a larger loop. The larger loop contains the primary
controller, also known the master controller. The unique characteristic of this control
14

configuration eliminates the need for disturbances to be explicitly measured, unlike the
feedforward control.
There are various disturbances to the setup: disturbance to the set point change, disturbance to
the inflow rate, and disturbance to the outflow rate. Graphical results of the disturbances are
presented.

4.2.1 Disturbances to Set Point Change


The reference level, or the setpoint, was increased from 7V to 8V within seconds of starting
the system. This is as seen from Figure 14 where there was a slight overshoot of the FT2
curve which adjusted the flow of the system. After the system has reached steady state, the
setpoint is decreased back to 7V and allowed to attain steady state. An undershoot was
observed but it was quickly adjusted back to the setpoint. The observations made were in
accordance to theory which stated that offsets in a cascade control system will be eliminated.

Set Point Change


160
140
120
100
80
60
40
Setpoint
Voltage (V)Level20
0
-20
-40

Level

F2-Ref

FT2

Time (s)

Figure 13. Plot of Responses when Set Point Change

It is noted that with the inclusion of F2-Ref, which is the input to the slave controller, the
accuracy of the graph is greatly reduced. Therefore, for the subsequent graphs, F2-Ref will be
removed.

15

Set Point Change (Magnified)


10
8
6
4
2
0
Voltage (V)

Time (s)
Level Setpoint

Change in
set-point
7V to 8V
8V to 7V

Level

FT2

Figure 14. Plot of Responses when Set Point Change (Magnified)


Time of
Time at steady
Time taken to reach
Change (s)
State (s)
steady state (s)
93.80
122.70
28.90
138.60
164.30
25.70

Maximum deviation in
level (V)
+0.110
-0.882

4.2.2 Disturbances to Inflow Rate


4.2.2.1 Inflow disturbance of 2.5V
The inflow of the system at steady state is decreased from 10V to 7.5V a set point deviation
by -2.5V. When the system has reached a new steady state at 7.5V, the inflow is increased
from 7.5V to 10V a set point deviation of +2.5V.

Inflow disturbance of +/- 2.5V (Magnified)


8
7
6
5
4
3
2
1
0
Voltage (V)

10V to 7.5V

Level SP

Level

7.5V to 10V

TF2

Time (s)

Inflow
disturbance
10V 7.5V
7.5V 10V

Time of
Time at steady State
Time taken to reach
Change (s)
(s)
steady state (s)
42.60
52.90
10.30
105.00
118.80
13.80
Figure 15. Plot of Responses with Inflow Disturbance of 2.5V

Maximum deviation in
level (V)
-0.026
-0.010

16

The inflow is perturbed at t=42.6s, where the inflow voltage is decreased from 10V to 7.5V.
The system took 10.6s to stabilize and reach a new steady state at t=52.9s. The inflow is then
perturbed again at t=105s, where the inflow voltage is now increased from 7.5V to 10V. It
took roughly same time for the controller to control the perturbation and bring the system to a
steady state, taking 13.8s for the system to stabilize. As seen from the graph, there are
minimal oscillations when the perturbations are introduced, hence the deviations of -0.026V
and -0.010V are rather small. F2-Ref graph also showed minimal deviation.
It is to be noted that cascade control is able to bring the system to a steady state within
seconds, while feedforward control took minutes to do so. This rapid response as compared to
feedforward control is one hallmark of cascade control, as the slave controller first minimized
effects of the disturbances before the effects of the disturbances reach other components in the
control loop. The slave controller can hence respond quickly to the disturbance in the inflow
by altering the power of the pump, hence altering the water level to reach steady state.

4.2.2.2 Inflow disturbance of 5V


The inflow of the system at steady state is decreased from 10V to 5V a set point deviation
by -5V. When the system has reached a new steady state at 5V, the inflow is increased from
5V to 10V a set point deviation of +5V.

Inflow disturbance of +/- 5.0V (Magnified)

Voltage (V)

8
7
6
5
4
3
2
1
0
Level Setpoint

10V to 5V

Level

5V to 10V

TF2

Time (s)

Inflow
disturbance
10V 5V
5V 10V

Time of
Time at steady
Time taken to reach
Change (s)
State (s)
steady state (s)
43.70
63.20
20.00
87.50
107.50
30.00
Figure 16. Plot of Responses with Inflow Disturbance of 5V

Maximum deviation in
level (V)
-0.017
-0.028

The inflow is perturbed at t=43.7s, where the inflow voltage is decreased from 10V to 5V.
The system took 29.5s to stabilize and reach a new steady state at t=63.2s. The inflow is then
perturbed again at t=87.5s, where the inflow voltage is now increased from 5V to 10V. It took
20s for the system to stabilize. From the graph above, it can be seen that the deviations are
now greater in magnitude as compared to that when the inflow disturbance of the system is
smaller at 2.5V instead of 5V. The maximum deviations are -0.017V and -0.028V.
17

As compared to the previous situation where the inflow disturbance to the system is 2.5V,
inflow disturbance of 5V results in a poorer controller action, as seen from an increase in the
magnitude of maximum deviations and the time taken to reach steady state.

4.2.2.3 Inflow disturbance of 7.5V

Inflow disturbance of +/- 7.5V (Magnified)

Voltage (V)

8
7
6
5
4
3
2
110V to 2.5V
0
Level Setpoint

Level

FT2

Time (s)

The inflow of the system at steady state is decreased from 10V to 2.5V a set point deviation
by -7.5V. When the system has reached a new steady state at 2.5V, the inflow is increased
from 2.5V to 10V a set point deviation of +7.5V.
Inflow
disturbance
10V 2.5V
2.5V 10V

Time of
Change (s)
116.00
567.60

Time at steady
State (s)
403.90
609.60

Time taken to reach


steady state (s)
287.90
42.00

Maximum deviation in
level (V)
-2.770
+0.189

Figure 17. Plot of Responses with Inflow Disturbance of 7.5V

The inflow is perturbed at t=116.0s, where the inflow voltage is decreased from 10V to 2.5V.
The system took 287.9s to attain a steady state, where the maximum deviation in level is
-2.77V. The inflow is then perturbed again at t=567.6s, where the inflow voltage is increased
from 2.5V to 10V. The system took a much lesser time (t=42s) to attain steady state this time
round, which the maximum deviation from set point is at a much smaller value at +0.189V.

18

The failure of the cascade controller to handle the perturbation could be due to the controller
reaching its saturation point, which will prevent the large magnitude of the perturbation from
being corrected. Similarly, F2-Ref plot showed large deviations in voltage with respect to
time, showing how the cascade controller is unable to handle such large perturbation to the
system.

4.2.3 Disturbances to Outflow Rate


4.2.3.1 Outlet Valve: Half-closed

Outlet Valve: Half-closed

Voltage (V)

8
7
6
5
4
3
2
1
0
Level Setpoint

Level

FT2

Time (s)

Figure 18. Plot of Responses with Outflow Valve Half-closed


Disturbance

Time of
Change (s)

Time at steady
state (s)

Time to reach
steady state (s)

Maximum deviation
in level (V)

Fully-opened to half-closed
Half-closed to fully-opened

153.50
252.90

185.60
320.00

32.10
67.10

+0.030
-0.090

Disturbance to the outflow rate was made by adjusting the outlet valve to half-closed. At
153.5s, the outlet valve was adjusted from fully-opened to half-closed. This change led to an
accumulation of water in the vessel as the outlet flow rate was decreased. In order to maintain
the water level in the vessel at its set-point, adjustments to the system was made by
decreasing the pump voltage as seen from the decrease of voltage of FT2. After approaching
steady state for a moment, the outlet valve was adjusted back to be fully-opened at t=252.9s.
This resulted in the increase of pump voltage as seen from the rise of the FT2 curve before
achieving steady state. However, the final voltage of FT2 at steady state was higher than the
initial
voltage
before
disturbance
was
present.

19

4.2.3.2 Outlet Valve: One-third-closed

Outlet Valve: One-Third Closed

Voltage (V)

8
7
6
5
4
3
2
1
0
Level Setpoint

Level

FT2

Time (s)

Figure 19. Plot of Responses with Outflow Valve One-third-closed

Disturbance

Time of
Change (s)

Time at steady
state (s)

Time to reach
steady state (s)

Maximum deviation
in level (V)

Fully-opened to one-third-closed
One-third-closed to fully-opened

118.00
167.60

139.00
185.70

21.00
18.10

+0.015
-0.029

Another disturbance was done to the outlet valve by closing it by one-third. As this
adjustment was smaller than before, the drop in voltage was also lower than before as seen
from the Figure above. The explanation for the change in voltage is similar to the explanation
as the previous change in outlet flow.

4.2.3.3 Outlet Valve: Two-third-closed

Outlet Valve: Two-third Closed

Voltage (V)

8
7
6
5
4
3
2
1
0
Level Setpoint

Level

FT2

Time (s)

Figure 20. Plot of Responses with Outflow Valve Two-third-closed

Disturbance

Time of
Change (s)

Time at
steady
state (s)

Time to reach steady


state (s)

Maximum deviation
in level (V)

Fully-opened to two-third-closed

118.80

214.50

95.70

+0.057
20

Two-third-closed to fully-opened

279.00

325.80

46.80

-0.116

Lastly, the outlet valve was closed two-third, the largest adjustment made to the outlet flow
rate. In this case, the drop in the pump voltage was the largest as seen from the drop in the
FT2 curve. From all the responses obtained with the changes in the outflow conditions, it can
be seen that the cascade system is able to provide a fast and effective response when sudden
changes were made to the outflow rate.

4.3 Discussion
4.3.1 Feedforward Control

Draw a block diagram of the system.


Figure 21. Block Diagram of Feedforward System

In what circumstances would you recommend a feedforward control system


rather than a cascade or single loop feedback control system? Point out any
limitations of the feedforward control system.

A feedforward control system is highly recommended to be used when the model of process
(process dynamic) is well investigated and known. For instances, the process dynamics
should be well understood which the variables present in the system can be measured and no
other unknown disturbances presented in the system. The feedforward controller will take
corrective action to the known disturbances before the process is upset. However, for both the
cascade and single loop feedback control system will only take corrective actions to the
system after the presented disturbance has affected the system. This results in a time lag in the
disturbance correction. Thus, the cascade and single loop feedback control system is
recommended when the disturbances present in the system is significant and unknown.
The limitation of the feedforward control system can be considered as the disturbances have
to be understood and measured, so that the corrective action can be taken before the process is
upset. Thus, the model of the system (process dynamics) has to be studied and known to
ensure accuracy in controlling the process. Besides, feedforward control system will only
correct the measured disturbance and it will not accounted for those unknown or not
21

measured disturbances in the system. To achieve a better performance, feedforward control


should be used together with a feedback control in order to account for all the disturbances
present in the system.

4.3.2 Cascade Control


1

Draw a block diagram of the system.

Figure 22. Block Diagram of Cascade System

What is the advantage of cascade control compared with single loop feedback
control? Point out limitations of cascade control system.

The primary difference between a cascade control and a single loop feedback loop is the
configuration in which cascade control has a feedback loop inside another larger feedback
loop. The advantage of a cascade control is only beneficial and feasible when the secondary
loop reacts at a much faster speed than the primary loop. In addition, the secondary loop has
to detect the greater disturbances present in the system. Hence, only in this configuration will
the cascade control be able to provide better control performance. The faster response of the
secondary loop will allow any disturbances present to be quickly adjusted and corrected
before it is experienced by the primary loop. Resultantly, the faster response will lead to a
reduction in lag time of the primary loop, therefore reducing the sensitivity to the upsets
present in the primary loop.
The main limitations of a cascade control system are high capital cost due to additional
instruments and equipment required. In addition, the cascade control system has a complex
control structure, which requires manual tuning of the controllers in the loop where
adjustments are made.

Look carefully at a single loop feedback control system and explain why, when
the valve is considered, it can be considered as a cascade control?

For a system to be considered a cascade control system there must be a presence of 2 control
loops: where the secondary (slave) control loop is nested within the primary (master)
control loop. The error signals from the master loop will then become the reference input to
the secondary loop.
22

Figure 23. Diagram of cascade control inside a feedback control system

From Figure 23, it depicts a primary level controller acting as a master control loop, analysing
the error signal and providing the reference set point for the second PID controller within the
valve to control the pump speed. The addition of the valve coupled with the second PID
controller actually forms a secondary loop within the bigger master loop, creating a cascade
control system, allowing a rapid response to the disturbances.

5. Error Analysis

1) When the drain valve for cascade control was adjusted manually, the size of
the step change was unable to be measured accurately. Therefore a
computerized flow indication control can be implemented instead of manual
opening. This will provide a much more accurate step change.

2) There are always tiny fluctuations in the graphs observed. This is probably due
to the noise affecting the system. A possible noise here is the splashing of the
water on the surface of the liquid when the water drips from the top of the
tank, which might affect the level indicators reading of the level in the tank.
Therefore to mitigate this noise, we can allow the water to flow from the
bottom instead of the top of the tank so that the surface of the liquid will not
have splashing and will have a smooth rise or fall of the level.

3) Time is a factor affecting the accuracy of the experiment. Since the experiment
takes a significant amount of time for every try, we were unable to conduct
each part of the experiment more than once. Ideally we would be able to
conduct the experiment repeatedly so that the average response time can be
taken. This is avoid random errors caused during the experiment.
23

6. Lab Safety Analysis

1) Since there is filling of the tank in the experiment, we need to make sure that
the level does not increase to the brim and overflow. The possible hazard of
overflowing might lead to wet electrical appliances such as pump which will
lead to electrical hazards. It can also create a wet environment which might be
susceptible to slipping.

2) In case of any anomaly, all the electrical appliances should be turned off such
as the pump and the computer.

3) All electrical appliances should be handed with dry hands

4) Before conducting the experiment, it should be ensured that all appliances are
working individually so as to not have any electrical hazards when the
experiment is conducted.
Generally the experiment is safe chemically because the only liquid used if water.
7. Conclusion

Feedforward Control

Cascade Control

Advantages

Immediate action to prevent the


measured disturbance from
affecting the system

Works well for significant


disturbances when used in
tandem with feedback control
Theoretically
perfect
control( prevents the measured
disturbance fully)

When manipulated variable is


affected by unknown disturbances.
Corrective action in response to any
unmeasured disturbance within the
slave loop which prevents the
controlled variable from deviating
from the setpoint
Useful when there is a secondary
process to which its manipulated
variable exists which can be
adjusted by the primary loop
24

Faster corrective response to


disturbance
(affecting
the
manipulated variable) involving the
slave loop before the master loop
control variable deviates from set
point- efficient for high occurrence
disturbance

Limitations

Accurate
required

Only applicable for measurable

disturbances

Cannot be used if the


disturbance is not measurable

process

model

Tuning is challenging and timeconsuming


The speed of the slave loop must be
ensured to be faster than the Master
loop to have effective cascade
control

8. References
Ponton J.W. (2007) Variations on Basic Feedback Control. Retrieved from
http://www.see.ed.ac.uk/~jwp/control06/controlcourse/restricted/course/second/course
/lecture6.html
Seaborg D.E., Edgar T. F., Mellichamp D.A., Doyle F.J. (2010). Process Dynamics
and Control (3rd Edition). Wiley.
Woolf P. (2007). University of Michigan Chemical Engineering Process Dynamics
and Controls Open Textbook. Retrieved from
https://controls.engin.umich.edu/wiki/index.php/Main_Page

25

You might also like