2015 International Conference on Advanced Computing and Communication Systems (ICACCS -2015), Jan.

05 – 07, 2015, Coimbatore, INDIA

Design and Implementation of Two Tank Conical Interacting System Using Intelligent
Technique (PSO)

Dr.K.Malarvizhi, Professor &Head of the Department, R.Kiruba, Assistant Professor,

Department of EIE, Department of EIE,
SNS College of Technology, SNS College of Technology,
Coimbatore, India. Coimbatore, India.
Email: malarkathir29@yahoo.co.in Email: kirubarajaram1790@gmail.com

Abstract— This paper deals with the modeling and liquid level and main goal is to track the set point of the liquid
control of highly Non linear system using intelligent technique level for the requirement given. In olden days, the traditional
namely Particle Swarm Optimization (PSO). Control of method called classical PID is used for controlling the process
process or process variable in process industries always variable To implement PID controller , three parameters
employ a non linear system i.e., Two Tank Interacting proportional gain , integral gain and derivative gain are to
Conical tank System (TTICS) is to be chosen due to its
determined very accurately and many approaches liken Ziegler
greater advantage. Conical tank system has varying cross
sectional area, and its level (process variable) is constantly Nichols , Cohen coon method were developed [6]. In these
changing and exhibits a non linear dynamic behavior and so traditional approaches the steady state is not achieved properly
controlling of this system becomes a challenging problem. So and it doesn’t track the set point accurately because correct
in this proposed work, the mathematical modeling for a tuning parameters are not identified. To overcome this
TTICS is obtained using real time process data. weakness, the new artificial techniques or intelligent
So for controlling the level of the process for proposed techniques like Genetic algorithm, ant colony optimization,
system, the process is tuned using the intelligent technique and particle swarm optimization are introduced in order to
called Particle Swarm Optimization (PSO) in which it is tune the process and tuned parameters are given to the non
guided by cost function, which considers the minimization of
linear system to control the process variable. Here in this
Integral Square Error (ISE), Integral Absolute Error (IAE)
and settling time. It is evident, from the simulation results proposed work, an attempt is made to control the process
that the control of level for the process is tracked and its variable for a TTICS using particle swarm optimization. In
effectiveness is compared based on the various performances fact, the PSO technique can produce highly optimized and
indices which has also been tuned using other tuning method solution within the shorter computation time and has stable
like Ziegler Nichols method. convergence characteristics [7]. Thus for this advantage, the
PSO technique has found attention towards wide applications.
Keywords—TTICS, Particle swarm optimization (PSO), Cost II. PROCESS DESCRIPTION
Function, ISE , IAE.
A. Mathematical Modeling of TTICS
I. INTRODUCTION Mathematical model is the one which gives complete
A system is said to be non linear when the description about the process under consideration using some
performance can’t be described by equations of first order mathematical concepts [14] [15]. The process that involves in
[14][15]. In most cases the process are non linear and its developing a mathematical model is termed as mathematical
mathematical modeling is unable to provide the general and modeling. Mathematical modeling of the liquid tank system is
accurate solutions. Comparing to linear systems, the non linear derived using the Total Mass Balance Equation. According to
systems are difficult to understand and solve them. But most that principle,
of the process industries use cylindrical, conical, spherical
tanks which are basically non linear due to its varying cross
sectional areas and its major goal is to control the process In TTICS, the process contains two identical conical
variable. Comparing to other conventional tanks, conical tanks tanks; the two identical pumps deliver the liquid flows Fin1 and
offers many advantage such as inexpensive, reduced product Fin2 to the tank 1 and tank 2 respectively through the two
losses. control valves MV1 and MV2 which acts as a manipulated
Liquid level control systems is the one which mainly variable [1] [2]. The two tanks under consideration should
involve to control the manipulated parameter of liquid level, have the same cross sectional area expect the outlet because it
which in industry have a wide range of applications in various has a non linear behavior. Our main objective in mathematical
automation fields. As for the industrial production process is modeling is to find the transfer function for the system that is
considered, there are many places that need to control the relation between output variable H2(s)

[978-1-4799-6438-3/15/$31.00 ©2015 IEEE]

 2015 International Conference on Advanced Computing and Communication Systems (ICACCS -2015), Jan. 05 – 07, 2015, Coimbatore, INDIA

(liquid height in Tank 2) and input variable Fin1(s) (Inlet flow

to tank 1). The main objective is to find the change in height
Where Kp is the proportional gain, Ti is the integral
of tank 1 and change in height of the liquid in tank 2 time constant and Td is the derivative time constant.
In simple form, the PID controller transfer function is
respectively.
According to the equation 1, the mathematical
modeling for TTICS is as follows,

A. Implementation of PID using Traditional Tuning Method

(Ziegler Nichols Method)
Tuning of PID involves the adjustment of Kp, Ki, Kd
to achieve the user defined ‘optimal’ character of the system.
In traditional methods, the PID controller is being tuned by
ZN method, CC method etc [6]. Here in this work, the Ziegler
Nichols tuning method is adopted to tune the PID controller
implemented for TTICS to evaluate the performance of the
PID PSO method is superior in all aspects like optimal
solution and best gain values of PID controller for the given
system and to achieve better response of time domain
specification parameters and steady state response when
compared to ZN tuned method. The performance estimation of
PID controller design attempts to minimize the errors
produced by certain anticipated errors. The error is defined as
the desired response in which the level has to meet and the
Fig. 1. Liquid level control of Two Tank Interacting Conical actual response of the system which is being tuned using some
System technique. The performance is based on some criterion like
For steady state condition, H1 = 45 cm and H2 =20 Integral of Absolute Error (IAE), Integral of Square Error
cm and valve co efficient ȕ1 = 5 ȕ2 =2, the transfer function is (ISE), and Integral of Time Weighted absolute Error (ITAE) is
obtained as shown in Table 1

TABLE 1 Performance Estimation PID Controller

Name of the Formula

Where
Criterion
H1 and H2 = height of the conical tank 1 & 2.
Integral of Absolute
dH/dt = change in height of the liquid level.
Error (IAE)
A = area of conical tank
Integral of Squared
‫ן‬ = control valve co efficient.
Error (ISE)
Integral of Time
III. PID TUNING Weighted Absolute
Error (ITAE)
PID is abbreviated as proportional, integral,
derivative controller. The individual P, I, D terms are
integrated to form the standard three term controller. The PID B. Intelligent technique (PSO) for PID Tuning
controller is used in most of the industrial process application PSO is a now recently become a popular for its high
due its greater advantage [19][4]. Each term has its own performance optimization which possess highly desirable
advantages and disadvantages. The proportional controller is attributes, including the fact algorithm is easy to understand
mainly responsible for process stability, but there may present and implement [8][9]. It requires less computational memory
offset. The integral term reduces offset error but when the and it has only few lines of coding. The PSO conducts search
integral gains increases the process leads to instability. The from the population of each particles which is considered as
derivative controller will reduce the oscillations that occurring individuals in GA. Each particle in the population has its own
due to larger integral gains and it will anticipate for future velocity vector and position vector to represent
errors. the possible solution.
The continuous form of a PID controller, with input Consider an optimization problem that needs an
e(.) and output upid(.) is generally given as: optimization of variables simultaneously. Initially a swarm of

[978-1-4799-6438-3/15/$31.00 ©2015 IEEE]

 2015 International Conference on Advanced Computing and Communication Systems (ICACCS -2015), Jan. 05 – 07, 2015, Coimbatore, INDIA

particles are defined, where each particle is assigned a random

position in N dimensional problem space so that each
particle’s position corresponds to candidate solution to the
given optimization problem. Then each particle fly in the
random position within the problem space with velocity of the
each particle. Each particle has three choices in evolution (1)
Insists on oneself (2) from the present state, it will move
towards the optimum solution and it will remember its own
best solution called as local best. (3) Move towards the best Fig. 2. Block Diagram of PSO – PID for Two Tank Interacting Conical
solution in the whole population or best position among the System
population called as global best. At each time step, each of the
IV. SIMULATION RESULTS AND DISCUSSION
particle positions is scored to obtain a fitness value based on
how it solves the problem. The simulation have performed using MATLAB
Simu link program for tuning of PSO-PID controller that is
used for level control of TTICS governed system with the
are the co efficient of inertia or weighing parameters. transfer function given in the equation 8. In order to
emphasize the performance of PID tuned using PSO
C. Implementatio of PSO – PID Tuning for TTICS technique, we also implement the PID Controller tuned using
The non linear two tank interacting conical system Ziegler Nichols method.
has a conventional PID controller and it is tuned using PSO
A. TTICS tuned for PSO- PID Controller
technique [3][10][11], its block diagram is shown in Fig. 2.
The PID controller using PSO algorithm was developed to Based on the trial and error method, the following PSO
improve the step transient response of the non linear system parameters are used for verifying the performance of PSO PID
(TTICS). The PSO algorithm was mainly utilized to determine controller:
three optimal controller parameters like Kp, Ki, Kd such that • The member of each individual is Kp, Ki, Kd.
the controlled system could obtain the good steady state • Population size =25
response. The controlled parameters composed an individual • Weighing factor = 0.9 & 0.4
K by K = [Kp Ki Kd] hence there are three members in the • Range of three controller parameter Kp= 0 to 1.5 ,
individual. Ki= 0 to 1, Kd= 0 to 1.
Optimization accuracy of the PSO technique mainly relies • Change in velocity = gain/2
on the Objective Function which is used for guiding the • Acceleration constant =2.
algorithm. In this work, the objective function is chosen as
Minimization problem i.e. minimization of errors namely IAE, After finding the controller values, the proposed
ISE and ts as controller settings are applied in simulation mode to study the
controller performance on the conical tank system with
different operating regions (15 cm and 25 cm). The simulation
Where P is the dimension of search population
result in Fig shows the step response of the TTICS for the
The searching procedure for the proposed PSO –PID
given set points (15 cm and 25 cm) tuned using PSO. It is also
for TTICS are as follows
evident from the Fig. 3 that PSO – PID controller will follow
Step 1: Select the upper and lower bounds of the three control
the changes of set point with small overshoot in the initial
parameters and initialize randomly the individuals of the
condition.
population including searching points, velocities, LBest,
GBest.
Step 2: For each individual K of the population, by using
some criterion calculate time domain specification.
Step 3: Calculate the value of each individual in the
population using the objective function
Step 4: compare the evaluation of each individual with the
LBest and the best evaluation among the GBest.
Step 5: Modify the value of velocity and position according to
the each individual.
Step 6: Stop the algorithm when it reaches the maximum
iterations or the desired value.
Fig. 3. Step response of TTICS for PSO – PID System at
Step 7: The individual that generates the latest GBest is called H=15 Cm and H = 25 Cm.
as optimal Parameters for the non linear system. It is also observed from the Fig. 3 that the PSO- PID
Step 8: The PSO algorithm continuously adjusts the gain of controller provides a smooth response with improved rise
PID controller until the f(p) is minimized. time, peak time, Overshoot time, settling time and also IAE,

[978-1-4799-6438-3/15/$31.00 ©2015 IEEE]

 2015 International Conference on Advanced Computing and Communication Systems (ICACCS -2015), Jan. 05 – 07, 2015, Coimbatore, INDIA

ISE values are also less compared to Ziegler Nichols tuned

PID controller.
TABLE 2 PID Tuning Values for various operating points

Set Point (in Tuning Parameters

Cm)
Kp Ki Kd

15 0.45 0.003 0.1

25 0.65 0.004 0.1

Fig. 4. Step response of TTICS for Ziegler Nichols – PID System at H=15
Cm and H = 25 Cm
The Table 5 shows the performance criterion values
TABLE 3 Performance Criteria of TTICS tuned using PSO PID for TTNICS tuned using ZN-PID controller, that the
Controller for various Set Points overshoot time is very high and rise time is very large
compared to PSO –PID method and the system has sluggish
Value response.

Specification H2=15 Cm H2=25Cm TABLE 5 Performance Criteria of TTICS tuned using ZN-PID Controller

Settling Time (ts) (Sec) 42.67 45.02 Specification Value

Rise Time (tr) (Sec) 20.56 22.45 H2=15 Cm H2=25 Cm

Maximum Overshoot 0 0 Settling Time (ts) (Sec) 1987.5 2325.5

(Mp) (Sec) Rise Time (tr) (Sec) 498 532.8

Maximum Overshoot 5 4
From the graph, it has seen that approximately around in (Mp) (Sec)
25 iterations the PSO method can prompt convergence and
obtain good evaluation value, thus achieve better performance
criterion that are rise time, settling time, percentage of Table 6 shows the performance comparison for
overshoot and steady state error condition. various time domain indices i.e. IAE and ISE error values, it
B. TTICS Tuned for Ziegler Nichols Method depicts that PSO method performs well compared to ZN
In order to emphasize the advantage of proposed method because the error values are very less in PSO tuned
PSO-PID controller, the Ziegler Nichols method has PID controller for TTICS
adopted[12]. The simulation results shown in Fig 4, depicts
the step response of the system for different set points (15 cm TABLE 6 Performance Indices Comparison
and 25 cm) tuned using Ziegler Nichols method. Compared
to PSO – PID method, the ZN method will follow the set Set point Controller ISE IAE
point with larger overshoot time in the initial condition. (Cm)
There will be larger error between the desired response to the
actual response of the system i.e., IAE, ISE values are larger. 15 PSO 145.6 173.2

TABLE 4 PSO - PID Tuning Values for various operating points ZN 152.5 181.7
when Ziegler Nichols Method
Tuning Parameters 25 PSO 153.5 184.6

Set Point (in Cm) Kp Ki Kd ZN 160.4 202.5

15 0.8 0.032 0

25 1.2 0.06 0

[978-1-4799-6438-3/15/$31.00 ©2015 IEEE]

 2015 International Conference on Advanced Computing and Communication Systems (ICACCS -2015), Jan. 05 – 07, 2015, Coimbatore, INDIA

V. CONCLUSION AND FUTURE WORK [4] Pedret, C., Vilanova, R., Moreno, R. and Serra. I. “A refinement
procedure for PID controller tuning”,Computer and Chemical
In this paper, to elucidate the performance of Engineering, vol.26, pp. 903-908, March 2002.
proposed work, initially the conventional PID controller was [5] Tan, K.K., Huang, S. and Ferdous, R, “Robust self tuning PID controller
tuned using Zeigler Nichols tuning for TTICS in order to for nonlinear systems” Journal of Process Control, vol.12, pp.753-761,
March 2002.
control the level in the system. The main aim PID - Ziegler
[6] Valerio, D. and Costa, J. S, “Tuning of fractional PID controllers with
Nichols controller tuning was done and the control of level in Ziegler-Nichols-type rules” Signal Processing, vol. 86, pp.2771-2784,
the TTICS has not achieved accurately, there exists larger April 2006.
error between the set point and the output tracked by the PID [7] Gaing, Z.L, “A Particle Swarm Optimization approach for optimum
controller when tuned by Ziegler Nichols Method. Secondly, design of PID controller in AVR system”, IEEE Transactions on
PSO PID controller has been used for tuning TTICS, in this Energy Conversion, vol.19, pp.384-391, April 2004.
methodology the set point has tracked accurately and there is [8] Kennedy, J. and Eberhart, R.C, “Particle swarm optimization”
Proceedings of the International Conference on Neural Networks,
no larger error and steady state performance and time domain pp.1942-1948, 1995.
parameters like settling time for PSO PID takes 0.02 sec to [9] Fan, H, “A modification to particle swarm optimization algorithm”,
settle in set point, in contrast PSO Ziegler Nichols takes 4.4 Engineering Computations, vol. 19,pp. 970-989, 2002.
sec to settle, rise time of the PSO PID is less and maximum [10] Asokan, P., Baskar, N., Babu, K., Prabhaharan, G. and Saravanan.R ,
overshoot time is instantaneously reduced in PID PSO “Optimization of surface grinding operations using particle swarm
method compared to PID Ziegler Nichols method and finally optimization technique”, Journal of Manufacturing Science and
Engineering, vol. 127, pp.885-892, 2005.
the control of level has achieved effectively.
[11] Liu, Y., Zhang, J. and Wang, S, “Optimization design based on PSO
Therefore, it is clear from the results that the proposed PID algorithm for PID controller”, Proceedings of 5th World
PSO for TTICS has more robust stability and efficiency and Congress on Intelligent Control and Automation, vol. 3, pp. 2419-
can able to solve the searching and tuning problems of PID 2422.2002.
controller parameters more easily and quickly than compared [12] Susmita Das, Ayan Chakraborty and Jayanta Kumar Ray , “Study on
Different Tuning Approach with Incorporation of Simulation
to Ziegler Nichols method. Aspect for Z-N (Ziegler- Nichols) Rules”, International Journal of
In future, work can be extended by implementing Scientific and research Publications, vol 2, ISSN: 2250- 3153.2012.
ANFIS as a system identifier for a non linear system which [13] Ziegler, J.G. and N.B. Nichols,(1942), “Optimum Settings for
uses hybrid learning method for better performance and to Automatic Controllers, Trans. ASME, vol. 64, pp.759- 768.
predict the output for a given system. [14] George Stephanapoulos , “ Chemical Process control”,Prentice Hall of
India,New Delhi,1990.
[15] Wayne Bequette.B, “Process control Modeling, Design and simulation”,
REFERENCES Prentice Hall, USA,2003.
[16] Astrom K.J and Hagglund T, “ PID Controllers: Theory,Design and
Tuning”, 2nd Edition, ISA Publication, North America,1995.
[1] Ravi V.R and Thyagarajan T,“A decentralized PID controller for
[17] Tan.W, Marquez and Chen .T, “Performance Assessment of PID
Interacting non linear System”, proceeding of emerging trends in
Controllers”, Journal of Control Instrumentation System, vol. 32, pp.
Electrical and Computer Technology, pp. 297- 302, April 2008.
158 – 166,2004.
[2] Nithya, S., G. Abhay Singh, N. Sivakumaran, T.K. Radhakrishnan, T.
[18] Yumana .M and Seborg D.E, “A new Method for Online Controller
Balasubramanian and N. Anantharaman, “ Design of Intelligent
Tuning”, Aiche Journal, Vol. 28, pp. 434-440,1982.
controllers for non linear Process”, IJSST, vol 9 no.4,pp 25-31,April
2008. [19] Skogestad .S, “ Probably the best PID tuning rules in the world”, Aiche
Journal, Vol.19,pp.276 – 280,2001.
[3] Wan Azhar Wan Yusoff, Nafrizuan Mat Yahya and Azlyna Senawi
“Tuning of Optimum PID Controller Parameter Using Particle Swarm [20] Soumya Ghosal, RajkumarDarbar, BiswarupNeogi, Achintya Das and
Optimization Algorithm Approach”, Fakulti Kejuruteraan Mekanikal, Dewaki N. Tibarewala (2012), “Application of Swarm Intelligence
Universiti Malaysia Pahang, 2006. Computation Techniques in PID Controller Tuning: A Review”,
Proceedings of the International Conference on Information Systems
Design and Intelligent Applications, vol 132, pp-195-208, Sep 2012.

[978-1-4799-6438-3/15/$31.00 ©2015 IEEE]