CHAPTER 5
HARDWARE RESULT
5.1 Introduction
In the project report, the open loop response of a conical tank, as well as its
closed loop response, will be showcased alongside the output of a backpropagation
Artificial Neural Network (ANN). The open loop response of the conical tank will serve
as a fundamental reference point, illustrating how the system behaves without any
external control or feedback mechanisms. This response will demonstrate the inherent
dynamics and characteristics of the tank system under varying conditions.
Subsequently, the closed loop response will be presented, highlighting the impact of
control strategies on system behaviour. By implementing feedback control, the closed
loop response aims to improve the stability, robustness, and performance of the
conical tank system. Finally, the output of the backpropagation ANN will be introduced,
offering a computational approach to modelling and controlling the system. The ANN's
ability to learn from training data and adapt to changing conditions will be explored,
showcasing its potential as an alternative or complementary control method to
traditional feedback control strategies. Through these comparative analyses, the
report aims to provide insights into the effectiveness and applicability of different
control approaches in real-world systems.
5.2 Hardware Implementation Kit
Figure 5.1 Conical Tank System
5.3 Open Loop Response
Front Panel Block Diagram
Figure 5.2 Open Loop in LabVIEW
Open Loop Response
40
35
30
25
Level (cm)
20
15
10
0
1101
1651
1
51
101
151
201
251
301
351
401
451
501
551
601
651
701
751
801
851
901
951
1001
1051
1151
1201
1251
1301
1351
1401
1451
1501
1551
1601
1701
Time (sec)
Set Point Sensor Readiing
Figure 5.3 Open Loop Response in Conical Tank
The parameters observed in the open-loop response graph are crucial for
calculating the transfer function of the conical tank system. By analysing the
relationship between the setpoint level and the corresponding reading level of the tank
over time, researchers can derive the transfer function that characterizes the dynamic
behaviour of the system. This transfer function encapsulates the system's input-output
relationship, describing how changes in the setpoint level propagate through the
system to affect the tank's actual reading. By obtaining the transfer function, engineers
and scientists gain deeper insights into the underlying dynamics of the conical tank
system, enabling them to design effective control strategies and optimize system
performance for various applications.
5.4 Closed Loop Design and Output Waveform
Figure 5.4 Closed Loop Block Diagram
Figure 5.5 Closed Loop Front Panel
12 cm
20
15
Level (cm)
10
0
3060
3725
1
134
267
400
533
666
799
932
1065
1198
1331
1464
1597
1730
1863
1996
2129
2262
2395
2528
2661
2794
2927
3193
3326
3459
3592
3858
3991
4124
4257
4390
4523
-5
Time (sec)
Set Point Sensor Reading
Figure 5.6 Cohen Coon response for 12 cm
The Cohen Coon response for the 12 cm setpoint exhibits a peak overshoot of
17.7157 cm, corresponding to a percentage overshoot of 47.6308%. The system
reaches its peak value in 24 seconds, and the settling time is 32 seconds.
24 cm
30
25
20
Level (cm)
15
10
0
653
2446
1
164
327
490
816
979
1142
1305
1468
1631
1794
1957
2120
2283
2609
2772
2935
3098
3261
3424
3587
3750
3913
4076
4239
4402
4565
4728
4891
5054
5217
5380
Time (sec)
Set Point Sensor Reading
Figure 5.7 Cohen Coon response for 24 cm
For the 24 cm setpoint, the Cohen Coon response indicates a peak overshoot
of 26.5782 cm, with a corresponding percentage overshoot of 10.7425%. The system
achieves its peak value in 60 seconds, and the settling time is 66 seconds.
36 cm
40
35
30
25
Level (cm)
20
15
10
0
4476
8056
11636
1
896
1791
2686
3581
5371
6266
7161
8951
9846
10741
12531
13426
14321
15216
16111
17006
17901
18796
19691
20586
21481
22376
23271
24166
25061
25956
26851
27746
28641
29536
Time (sec)
Set Point Sensor Reading
Figure 5.8 Cohen Coon response for 36 cm
The Cohen Coon response for the 36 cm setpoint shows a peak
overshoot of 36.8159 cm, indicating a deviation from the desired level. The percentage
overshoot, calculated as 2.8664%, signifies the extent of this deviation relative to the
setpoint. It takes approximately 325 seconds for the system to reach its peak level,
and the settling time, the duration for the system to stabilize within an acceptable range
of the setpoint, is observed to be 525 seconds.
5.5 ANN Prediction and Response
Level
S.
h Transfer
No. Kp Ti
(cm) function
13.17
1 12 𝑒 −2.5𝑠 2.8066 66.667
10.5𝑠 + 1
8.8
2 24 𝑒 −3𝑠 12.5279 9.459
111𝑠 + 1
10
3 36 𝑒 −29.5𝑠 0.839 83.3333
271.5𝑠 + 1
Table 5.1 ANN Prediction for Kp and Ti values
12 cm
20
18
16
14
12
Level (cm)
10
8
6
4
2
0
1600
3691
1
124
247
370
493
616
739
862
985
1108
1231
1354
1477
1723
1846
1969
2092
2215
2338
2461
2584
2707
2830
2953
3076
3199
3322
3445
3568
3814
3937
-2
Time (sec)
Set Point Sensor Reading
Figure 5.9 ANN response for 12 cm
The response of the artificial neural network (ANN) model for the 12 cm setpoint
exhibits a peak overshoot of 17.2573 cm, corresponding to a percentage overshoot of
43.8108%. The peak time, indicating the time taken to reach the peak level, is
observed to be 18 seconds. Subsequently, the settling time, representing the duration
for the system to stabilize around the desired setpoint, is measured at 30 seconds.
24 cm
30
25
20
Level (cm)
15
10
0
691
1381
2071
2761
1
231
461
921
1151
1611
1841
2301
2531
2991
3221
3451
3681
3911
4141
4371
4601
4831
5061
5291
5521
5751
5981
6211
6441
6671
6901
7131
7361
7591
-5
Time (sec)
Set Point Sensor Reading
Figure 5.10 ANN response for 24 cm
The response of the artificial neural network (ANN) model for the 24 cm setpoint
demonstrates a peak overshoot of 27.984 cm, equivalent to a percentage overshoot
of 16.6%. The peak time, signifying the duration taken to reach the maximum
overshoot, is recorded at 64 seconds. Subsequently, the settling time, indicative of the
time required for the system to stabilize around the desired setpoint, is observed to be
72 seconds.
36 cm
40
35
30
25
Level (cm)
20
15
10
5
0 11509
23976
1
960
1919
2878
3837
4796
5755
6714
7673
8632
9591
10550
12468
13427
14386
15345
16304
17263
18222
19181
20140
21099
22058
23017
24935
25894
26853
27812
28771
29730
30689
31648
-5
Time (sec)
Set Point Sensor Reading
Figure 5.11 ANN response for 36 cm
The ANN model's performance for the 36 cm setpoint showcases a peak
overshoot of 37.6716 cm, indicating a modest overshoot percentage of 4.6433%. It
achieves peak response at 300 seconds, demonstrating its ability to reach the
maximum deviation from the setpoint. Moreover, the settling time for the system to
stabilize around the desired level is observed to be 315 seconds, highlighting the
model's dynamic response characteristics.
5.7 Conclusion
In conclusion, the comparison between the closed-loop response of a conical
tank system using conventional PI control and an artificial neural network (ANN) model
reveals interesting insights into their performance. While both control strategies aim to
regulate the tank's liquid level to a desired setpoint, they exhibit distinct characteristics.
The PI controller, tuned using the Cohen-Coon method, demonstrates satisfactory
performance with acceptable overshoot and settling times across different setpoint
values. On the other hand, the ANN model, trained using backpropagation, offers
competitive performance, albeit with slightly different response dynamics. The ANN's
ability to approximate complex nonlinear relationships allows it to adapt to varying
setpoints effectively. Overall, the study highlights the efficacy of both control
approaches and underscores the importance of selecting the appropriate method
based on the specific requirements and dynamics of the system.
