GOVERNMENT COLLEGE OF TECHNOLOGY

COIMBATORE – 641013
DEPARTMENT OF ELECTRONICS AND INSTRUMENTATION ENGINEERING
18NEE803 – PROJECT WORK
PROJECT REVIEW - 26/03/2024
MODELLING AND IMPLEMENTATION OF AI
BASED CONTROLLER FOR CONICAL TANK
1

Presented by,.
Guided by,.
DHARUN BALAJI P - 2016108
Dr.B.Achiammal M.E., Ph.D., DIVYA R M - 2016110
Assistant Professor of EIE SARAVANAKUMAR A - 2016127
HARIKISHAN V - 2016L09
 ABSTARCT
Industries face challenges in process control due to difficulties in user interaction
and data comprehension, especially in managing variables like those in Artificial
Intelligent (AI) based controllers for conical tank. The current interfaces create
obstacles in smoothly controlling processes. This project addresses these issues by
integrating an AI system, aiming to simplify variable control. It also focuses on
improving data presentation, crucial for understanding complex information related to
2

level control. In real-time implementation using advanced AI tools like LabVIEW and
MATLAB. This initiative aims to overcoming drawbacks in PI controllers, and
ensuring smoother operations and better decision-making
2
 PROBLEM STATEMENT
The problem at hand centers around the modelling and implementation challenges of an Artificial
Intelligence (AI) based controller for a conical tank within the realm of process control. Current
methodologies encounter difficulties in effectively representing the dynamic behavior of the conical
tank, incorporating the complexities of fluid dynamics and control parameters. Additionally, the
implementation of AI-based control systems demands a comprehensive understanding of real-time
execution, data acquisition, and seamless integration with existing tools like LabVIEW and MATLAB.
Addressing these challenges is crucial for optimizing control strategies, ensuring the efficient operation
3

of the conical tank, and advancing the capabilities of AI-driven controllers in the domain of process
control. The problem statement seeks to explore innovative solutions that bridge the gap between
theoretical modelling complexities and the practical implementation challenges in realizing an AI-
based controller for a conical tank. 3
 METHODOLOGY

4 DAQ

Conical
Tank
Personal Computer 4
MATHEMATICAL MODELLING OF THE CONICAL TANK

H = 60 cm
5 R = 31 cm

5
 MATHEMATICAL MODELLING OF THE CONICAL TANK
Using Balance equation, according to Newton’s Law:
Fin – Fout = ----(4)
tan θ = = = According to Bernoulli’s Theorem:
r = ----(1) Fout = k ----(5)
Volume of the tank : V = πr2h ----(2) Substituting Eqn. (3),(4) in Eqn. (5):
Substituting Eqn.(1) in Eqn.(2) : Fin – k = π()2
Simplifying: = = -
V= π)h 2 3

Let α= and β = kα
Differentiating with respect to time: 6

Then = - β
= π()2 ----(3)

6
 MATHEMATICAL MODELLING OF THE CONICAL TANK
Using Taylor’s series and linearizing the above differential equation, the transfer function of the system is obtained as :
T.F =
Where : C = and

S.No. Level Valve Coefficient Transfer function

h (cm) K
1 12 0.5

2 12 1

3 24 7 0.5

4 24 1

List of transfer functions for various heights and valve coefficients 7

 MATHEMATICAL MODELLING OF THE CONICAL TANK
S.No. Level Valve Coefficient Transfer function

h (cm) K
5 36 0.5

6 36 1

7 48 0.5

8 48 1

9 60 0.5

10 60 1

List of transfer functions for various heights and valve coefficients

8
 DIRECT SYNTHESIS TUNING METHOD
Consider a closed-loop system with unity feedback. The system parameters are as follows:
Gc(s) = Transfer Function of the Controller
Gp(s) = Transfer Function of the Process
R(s) = Input to the system
Y(s) = Output of the system

The overall Closed-Loop Transfer Function of the system is given as:

GCL(s) ==
Gc =
Let Gp(s) = and GCL(s) = 9

Then GC = + ------(1)
In the ideal form of a PI controller: Gc = Kc(1 + ) ------(2)
Equating Equations (1) and (2), the tuning parameters are obtained as : K c = , i =

9
 DIRECT SYNTHESIS TUNING METHOD
S.No. Level h Valve Coefficient Transfer function
Kp​ Ti
(cm) K
(min)​
1 12 0.5
0.079 17.68

2 12 1
0.158 8.84

3 24 0.5
0.05 100.014

4 24 1
0.111 50

5 36 0.5
0.046 275.6

6 36 1
0.0912 137.8
10

7 48 0.5
0.079 282.883

8 48 1
0.0158 1414.296

9 60 0.5
0.0707 494.176

10 60 1
0.0141 2470.67
10
 DIRECT SYNTHESIS TUNING METHOD
Graphs:
K=0.5, h=12 cm

11
 DIRECT SYNTHESIS TUNING METHOD
Graphs:
K=1, h=12 cm

12
 DIRECT SYNTHESIS TUNING METHOD
Performance Analysis:

Parameter DS-PI ANN-PI

VALVE COEFFICIENT K=0.5 K=1 K=0.5 K=1

RISE TIME (SEC) 864.5 1608.8 106.2 431.3

PEAK OVERSHOOT (%) 13.8 13.7 13.05 13.6

PEAK TIME (SEC) 1856.9 3456.9 236.9 877.78

SETTLING TIME (SEC) 3147.3 5989.9 384.6 1582.4

13
OPEN-LOOP LabVIEW DIAGRAM

14

14
OPEN-LOOP RESPONSE OF THE CONICAL TANK
SYSTEM FOR VARIOUS SETPOINTS

15

15
 5
TWO – POINT METHOD

16

16
 TWO – POINT METHOD
Level τ = 1.5(t2 - t1) Maximum level Minimum level K = Maximum level TF =
S.No t1 t2 td = t 2 - τ
cm s s s s cm cm - Minimum level

cm

1 12 6 13 10.5 2.5 14.079 0.907 13.17

2 24 40 114 3 111 22.76 13.96 8.8

3 36 120 301 271.5 29.5 33.73 23.73 10

17

4 48 147 359 318 41 39.08 35.01 4.07

5 60 18 25 10.5 14.5 40.88 38.99 1.89

17
CLOSED-LOOP BLOCK DIAGRAM IN LabVIEW

18
CLOSED-LOOP FRONT PANEL DIAGRAM IN
LabVIEW

19
OPEN-LOOP RESPONSE OF THE CONICAL TANK
SYSTEM FOR VARIOUS SETPOINTS

Setpoint = 12 cm Setpoint = 24 cm

Setpoint = 36 cm

20
FLOWCHART OF BACKPROPAGATION ANN

21

21
DEVELOPMENT OF BACKPROPAGATION ANN

22

22
 AI CONTROLLER OUTPUT

Setpoint = 12 cm Setpoint = 24 cm

Setpoint = 36 cm

23
 REFERENCES
[1] Tate Bestwick , Kyle V. Camarda, “Artificial Neural Network-Based Real-Time PID
Controller Tuning”, 2023, Vol.52, Pg.1609-1614
[2] Dhaya Sindhu Battina, “Application Research of Artificial Intelligence in Electrical
Automation Control”, International Journal of Creative Research Thoughts, 2015, Vol. 3, Issue
4, December 2015, ISSN: 2320-2882, Pg.784-787
[3] Varanon Uraikul, Christine W. Chan , Paitoon Tontiwachwuthikul, “Artificial intelligence
for monitoring and supervisory control of process systems”, Engineering Applications of
Artificial Intelligence 20, 2007, Pg.115–131
[4] Julius Schöning, Adrian Riechman, Hans-Jürgen Pfisterer, “AI for Closed-Loop Control
Systems: New Opportunities for Modeling, Designing, and Tuning Control Systems”, 14th
International Conference on Machine Learning and Computing, February 2022, ICMLC '22:
Pg.318–323
[5] M. L. Reddy, V. Indragandhi, B. Kushal and R. Rajasingh, “Integration of MATLAB and
LabVIEW for Motor Control Test Bench with Power Analysis”, Innovations in Power and
Advanced Computing Technologies (i-PACT), Vellore, India, 2019, Pg.1-6
24
 REFERENCES
[6] L. Ying, “The Simulation Design of PID Control Based on LabVIEW and MATLAB”,
International Symposium in Sensing and Instrumentation in IoT Era (ISSI), Shanghai, China,
2018, Pg.1-8
[7] P. Karas and Š. Kozák, “Highly nonlinear process model using optimal artificial neural
network”, Cybernetics & Informatics (K&I), Lazy pod Makytou, Slovakia, 2018, Pg.1-7
[8] Y. Chang and H. Huang, “Robust adaptive output sampled control of uncertain system with
unknown time-varying control coefficient based on characteristic model”, 36th Chinese
Control Conference (CCC), Dalian, China, 2017, Pg.1367-1372
[9] Dr.Rajeshkumar, Smriti Srivastava, “Artificial Neural Network based PID controller for
online control of dynamical systems”, IEEE 1st International Conference on Power
Electronics, Intelligent Control and Energy Systems (ICPEICES), 2016, Pg.1-10
[10] S. Vergura, E. Natangelo “Labview-Matlab integration for analyzing energy data of PV
plants”, International Conference on Renewable Energies and Power Quality (ICREPQ’10),
Granada (Spain), 23th to 25th March, 2010, Pg.1248-1252
25
Thank You!

26