International Journal of Engineering and Technical Research (IJETR)

ISSN: 2321-0869, Volume-3, Issue-4, April 2015

Design and Simulation of PSO Based PI Controller

for a Non-Linear Two-Tank Interacting Level
Process
R.Soundariya, Dr.S.Arulselvi
valves, six process tanks, two overhead tanks, two differential
Abstract In this work, a non linear two-tank interacting pressure transmitters, five level transmitters and rotameters
level process is taken-up for study. The mathematical model of .Instrumentation panel consists of two PID controllers, main
the two-tank interacting process is derived. A conventional PI power supply switch, pump switches, motorized control valve
controller is designed based on process reaction curve with ZN
switches and auxiliary switches for individual components.
technique. To improve the performance of PI controller, a
Particle Swarm Optimization (PSO) based PI controller PSO-PI
is proposed and implemented. The servo and regulatory III. WORKING PRINCIPLES
responses with PI and PSO-PI are obtained and discussed. From
the results and performance measures, it is observed that
PSO-PI is producing better results than conventional PI Fluid level in the tank is measured by level transmitter (LT).
controller. Output of LT is given to the data acquisition setup. It consists
of ADC and DAC. The differential pressure level transmitter
Index Terms Two-tank interacting process, PI controller (DPLT) measures the flow by sensing the difference in level
and PSO-PI. between the tanks. The DPLT then transmits a current signal
(4-20mA) to the I/V converter. The output of the I/V
I. INTRODUCTION converter is given to the interfacing hardware associated with
The process industries require the liquids to be pumped, the personal computer (PC). Control algorithms are
stored in tanks and then pumped to another tank. Many times implemented in Lab view software. It compares and takes
the liquid will be processed by chemical or mixing treatment corrective action on the motorized control valve. Based on the
in the tanks, but always the level of the fluid in the tanks must valve opening flow rate is manipulated. Rotameter can
be controlled. A level that is too high may upset the reaction visualize the flow rate.
equilibrium, cause damage to the equipment or result spillage
of valuable or hazardous material [1]. If the level is too low, it
may have bad consequences for the sequential operations.
Hence, control of liquid level is an important and common in
process industries. Conventional controllers are widely used
in industries since they are simple robust and familiar to the
field of operator. The most basic and pervasive controller
algorithm used in the feedback control is proportional integral
controller algorithm. PI controller is widely used control
strategy to control most of industrial automation process
because of its remarkable efficiency and simplicity. Hence, a
mathematical model is derived and a simulation is carried out
for the given mathematical equation. The conventional PI
controller parameters are designed based on
Ziegler-Nicholas. However, it will not give satisfactory
response for change in operating point. Hence, the Particle Fig.1. Experimental setup of a two-tank interacting process.
Swarm Optimization technique is proposed and implemented
to optimize PI controller parameters [4]. In this work, a The controller compares the controlled variable against set
simple performance criterion in time domain is proposed for point and generates manipulated variable as current signal
evaluating the performance of a PSO-PI controller that is (4-20mA) [2]. Here the controlled variable is the level (h2)
applied to a interacting two-tank process. and the manipulated variable is the flow rate
(qin).Specification dimension are tabulated in Table I and
II. PROCESS DESCRIPTION Table II ,respectively.
Fig.1 shows the photograph of the laboratory level process TABLE I
station. It consists of three pumps, two motorized control
SPECIFICATION OF TWO-TANK INTERACTING
PROCESS
Manuscript received April 18, 2015.
R.Soundariya, Department of Process Control and Instrumentation, COMPONENTS SPECIFICATIONS
Annamalai University, Tamil Nadu, India.
Dr.S.Arulselvi, Department of Process Control and Instrumentation,
Annamalai University, Tamil Nadu, India.

274 www.erpublication.org
 Design and Simulation of PSO Based PI Controller for a Non-Linear Two-Tank Interacting Level Process

MOTORISED Flow rate 50LPH tank2 (cm), respectively Both tanks have the same cross
CONTROL Characteristics Equal% sectional area A1(cm2) and A2(cm2), qL1 is the inflow of tank1
VALVE Valve action motorized as load disturbance (cm3/min) and qL2 is the inflow of tank2
control as load disturbance (cm3/min) [1and 2].
ROTAMETER Type variable area
qin(mv) qL1 (Load) qL2 (Load)
Range (0-100)LPH
float material i.e (0-1666
3
/min)
SS 316
PUMP RPM 4500
Discharge 1000(LPH)
Voltage 220/230 volts R1 R2
AC&DC h1
h2
PROCESS TANK Capacity 3 litres
Height 300 mm
Diameter 120 mm q1
A1 A2
LEVEL Input 24V DC
TRANSMITTER Height 0-400 mm WC
tank1 tank2
Type F capacitance

Fig.2.Two-tank interacting process

DIFFERENTIAL Input 0-400mm H2O
PRESSURE Supply 24V DC
TRANSMITTER Output 4-20mA at 24V Mass balance equation for tank 1
DC
dh1
A1 qin q1 (1)
TABLE II dt
DIMENSIONS AND VARIABLES FOR TANK1 AND
TANK 2. Assuming non -linear resistance to flow :

Parameter Dimension c a
q1 d 1 2 g (h h )
Area(A1,A2) 113.0973 cm2 A 1 2 (2)
1
Height(max) 25 cm Mass balance equation for tank 2
Diameter 12 cm
dh
0-1666 cm3/min A2 2 q q (3)
Inflow rate(MV),qin 1 o
dt
Process Variable (PV),h2 0-25 cm Assuming non-linear resistance to flow:

0-500 cm3/min
c a
q d 2
Flow rate (LV),qL1,qL2
2 gh
G Gravity (9.81N/m2) o A 2
2
a1 and a2 Area of the pipe outlet
(a1=3.5735cm2 , dh c a c a
d 1 (4)
a2=3.9012cm2) A 2
2 dt 2 g (h h ) d 2 2 gh
cd Discharge co-efficient A 1 2 A 2
1 2
( cd=0.08)

MV- manipulated variable, PV-process variable LV- IV. CONTROLLER TUNING

B. NonLoad
linearvariable
mathematical modelling of a two-tank A. Process reaction curve for two-tank interacting process
interacting level process
Fig.2 shows the process consists of two interacting liquid Fig.3 shows the simulated open loop response of
tanks. The volumetric flow into the tank1 is qin(cm3/min) , the interacting process. The level (h2) changes from 0 to 8.5 cm,
volumetric flow rate from tank1 to tank2 is q1(cm3/min), and when applying a step input in qin(50*16.66cm3/min) also the
the volumetric flow rate from tank2 is qo(cm3/min) [2]. The level (h1) changes from 0 to 18 cm due to interaction. The
height of the liquid level is h1(cm) in tank1 and h2(cm) in

275 www.erpublication.org
 International Journal of Engineering and Technical Research (IJETR)
ISSN: 2321-0869, Volume-3, Issue-4, April 2015
simulated process reaction curve(PRC) of h2 for step change INTERACTING PROCESS USING Z-N METHOD.
in qin for 499.8cm3/min is shown in Fig.4.
Mode
2 h1
PI 7.9351 56.61
h2
2
Level (cm)

1
C. Particle swarm optimization (pso) algorithm
Particle swarm optimization (PSO) algorithm, which is
1 used for optimizing the PSO difficult numerical functions and
based on metaphor of human social interaction, is capable of
5 mimicking the ability of human societies to process
knowledge [4]. It has roots in two main component
0 methodologies: artificial life (such as bird flocking, fish
0 50 100 1500
Time(sec) schooling and swarming) and evolutionary computation.
Although the PSO algorithm is initially developed as a tool
Fig.3. Simulated open loop response of h1 and h2 of for modelling social behaviour, it has been applied in
interacting Process. different areas.
R(s) E(s) U(s)
10

8
h2(cm)

6
Fig.5 . PSO based PI controller to control the level of

5 two- tank interacting process.

1000 2000 3000 4000 5000
Time(sec)
The block diagram of PSO based PI controller is shown in
Fig.4. Simulated PRC of h2 for step change in qin for Fig 5. Also the flow chart to optimize the PI controller is
499.8cm3/min. shown in Fig 6.

From the fig.4 the transfer functions are obtained and

tabulated in Table III. From the average transfer function, the
controller parameters are obtained using Z-N tuning rule [2].
For two-tank interacting process the PI controller parameters
are tabulated in Table IV.
TABLE III
TRANSFER FUNCTION MODEL OF TWO-TANK
INTERACTING PROCESS
Step Transfer Function Average Transfer
Input Function

Positive Step
Input
Negative Step
Input
Fig 6.Flow chart PSO-PI
A. Z-N tuning technique The equation for position and velocity
Based on the average transfer functionthe value of kc and
ti are calculated using Z-Ntechniques.The values of kc and ti Vn+1 = vn+c1rand1( )*(pbest,n-current position)+ c2rand1(
are tabulated in Table IV. )*(gbest,n-current position) (5)
TABLE IV
x (t +1) = xid (t) + vid(t +1) (6)
PI CONTROLLER SETTINGS FOR TWO-TANK

276 www.erpublication.org
 Design and Simulation of PSO Based PI Controller for a Non-Linear Two-Tank Interacting Level Process

performances than PI controller. .

D. The PSO learning algorithm is represented by
following steps 10
Step1 Select the number of iterations (n) and the PSO
learning rate (c1, c2).
8
Step2 Randomly generate position vector X and
associated velocity V of all particles in the 6

h2(cm)
population.
PSO-PI
Step3 Calculate each individuals fitness value, and 4
PI
then compare each individuals evaluation value
with best global particle value gbest the personal
best value pbest . Finally select the new best
2
value (gbest and pbest)
0
Step4 For every particle, update its own velocity and 0 2000 4000 6000 8000
position value. Time(sec)
Step5 If iteration=n, then go to exit, otherwise go to Fig.7. Servo response of h2 with PI and PSO-PI
step 3. controllers.
Step6 The best particles value will be selected as the
finial parameter set to form desired PI
controller.
25

20
During simulation the values for PSO algorithm
considered are tabulated in Table V.
15
h1(cm)

TABLE V
PARAMETER OF PSO ALGORITHM
Parameter Values 10 PSO-PI
Number of Iteration 20
Dimension 3 PI
Swarm Size 90 5
Correction Factor C1 1.2
C2 1.4
0
0 2000 4000 6000 8000
The optimized PSO-PI parameter using equation 5 and 6 Time(sec)
are tabulated
TABLE VI Fig.8.Servo response of h1 with PI and PSO-PI controllers.
PARAMETERS OF PI AND PSO-PI CONTROLLERS 600
Mode

PI 7.9351 56.61
qin(cm /min)

PSO-PI 2.03791 0.1005 400

PSO-PI
3

PI
V. SIMULATION RESULTS
200
A. Servo responses of level h2 with PI and PSO-PI
controllers.

The servo responses for two tank level interacting process 0

for step change in h2 from 6 to 8 cm, from 8 to 6cm is shown 0 2000 4000 6000 8000
in fig.7 by implementing PI and PSO-PI. The corresponding time(sec)
change in h1 and inlet flowrate qin are also shown in fig.8 and
fig.9, respectively. The performance measures for step change Fig 9. Response of PI and PSO-PI output qin for
in h2 from 6 to 8 cm is tabulated in Table VII. From the results servo response.
and Table VII, it is observed that the PSO-PI gives better

277 www.erpublication.org
 International Journal of Engineering and Technical Research (IJETR)
ISSN: 2321-0869, Volume-3, Issue-4, April 2015

TABLE VII 420 PSO-PI

COMPARISON OF PERFORMANCE MEASURES PI
LEVEL WITH PI AND PSO-PI FOR SERVO

qin(cm /min)
400
RESPONSE (6-8) CM

3
380
Servo response for h2 (6-8)cm
ts(sec) %overshoot ISE
360
PI 1988 12.825 353.4
PSO-PI 978 33.66 254.2 340
4000 4500 5000 5500
B. Regulatory response for Positive step change load time(sec)
variable qL1(+) Fig 12. Response of PI and PSO output(qin) for load
variation in +10% from qL1
A sudden load disturbance of +10% is given in inlet
flow rate of tank1 at 4000 sample from q L1(+) . Due to this C. Regulatory response for negative step change load
level in h1 increases for PI from 13.3 to 14.5 cm and PSO it variable qL1(-)
increases from 13.3 to 15.4 as shown in fig.11. The level h 2
also increases for PI from 6 to 6.5 cm and PSO it increases
from 6 to 6.9cm as shown in fig 10. The PI and PSO-PI A sudden load disturbance of -10% is given in inlet flow
Controller takes necessary action to reduce the flowrate q in rate of tank1 at 4000 sample from qL1(-) . Due to this level in
from 420 to 350 (ref fig12) In order bring back the level h 2 h1 decreases for PI from 13.3 to 12.1 cm and PSO it
back to 6cm as shown in fig 10. decreases from 13.3 to 11.1 as shown in fig 14. The level h 2
also decrease for PI from 6 to 5.5 cm and PSO it decreases
from 6 to 5.1 cm as shown in fig 13. The PI and PSO-PI
7 Controller takes necessary action to increase the flowrate q in
PSO-PI from 425 to 500 (ref fig 15) in order bring back the level h 2
PI back to 6cm as shown in fig 13.

6.5
h2(cm)

6.2
6
h2(cm)

6 5.8 PSO-PI
5.6 PI
5.4
4000 4500 5000 5500
time(sec) 5.2
Fig 10. Regulatory response of h2 with PI and PSO due
to load variation in +10% from qL1.
4000 4500 5000 5500
time(sec)
Fig 13. Regulatory response of h2 with PI and PSO-PI due to
load variation in -10% from qL1
PSO-PI
15 PI
14 PSO-PI
h1(cm)

14 13.5 PI

13
h1(cm)

13 12.5
12
12 11.5
4000 4500 5000 5500
time(sec) 4000 4500 5000 5500
Fig 11. Regulatory response of h1 with PI and PSO-PI due to time(sec)
Load variation in +10% from qL1.
Fig 14. Regulatory response of h1 with PI and PSO-PI due
to load variation in -10% from qL1.

278 www.erpublication.org
 Design and Simulation of PSO Based PI Controller for a Non-Linear Two-Tank Interacting Level Process

13.5 PSO-PI
500 PSO-PI
PI 13 PI
qin(cm min)

480 12.5

h1(cm)
12
3

460
11.5
440
11
420 4000 4500 5000
time(sec)
4000 4500 5000 5500 . Fig 17. Regulatory response of h1 with PI and PSO-PI
time(sec) due to load variation in +10% from q L2 .
Fig 15. Response of PI and PSO-PI output (qin) for load
variation in -10% from qL1.
500
TABLE VIII
PSO-PI
COMPARISON OF PERFORMANCE MEASURE OF PI

qin(cm /min)
LEVEL WITH PI AND PSO FOR REGULATORY 400
RESPONSE. 3
Regulatory Response h2 (6-8)cm 300
Controller +10% from qL1 -10% from qL1
ts(sec) ISE ts(sec) ISE 200
PSO-PI 1032 2822 1046 2821 4000 4200 4400 4600 4800 5000
PI 1460 9316 1275 9316 time(sec)
Fig 18. Regulatory response of h1 with PI and PSO-PI
due to load variation in +10% from q L2 .
D. Regulatory response for positive step change load
variable qL2(+) E. Regulatory response for Negative step change
load variable qL2(-)
A sudden load disturbance of +10% is given in inlet flow
rate of tank1 at 4000 sample from qL2(+) . Due to this level in A sudden load disturbance of -10% is given in inlet flow
h1 increases for PI from 13.3 to 13.5 cm and PSO it increases rate of tank1 at 4000 sample from qL2(-) . Due to this level in
from 13.3 to 13.6cm as shown in fig 17. The level h 2 also h1 increases for PI from 13.3 to 15.6 cm and PSO it increases
increases for PI from 6 to 7 cm and PSO it increases from 6 to from 13.3 to 15 as shown in fig 20. The level h2 also decreases
6.9cm as shown in fig 16. The PI and PSO-PI Controller for PI from 6 to 5.6 cm and PSO it decreases from 6 to 5.8cm
takes necessary action to reduce the flowrate qin from 420 to as shown in fig 19. The PI and PSO-PI Controller takes
200 (ref fig18) in order bring back the level h2 back to 6cm as necessary action to increase the flowrate qin from 420 to 750
shown in fig 16. (ref fig21) in order bring back the level h2 back to 6cm as
shown in fig 19.
7 PSO-PI
PI
6.5
h2(cm)

6
h2(cm)

PSO-PI
6 PI
5.5

5.5
4000 4500 5000 5
time(sec) 4000 4500 5000 5500
Fig 16. Regulatory response of h2 with PI and PSO-PI time(sec)
due to load variation in +10% from Fig 19. Regulatory response of h2 with PI and PSO-PI due
qL2 to load variation in -10% from qL2.

279 www.erpublication.org
 International Journal of Engineering and Technical Research (IJETR)
ISSN: 2321-0869, Volume-3, Issue-4, April 2015

16 REFERENCES
PSO-PI
PI [1] V. Rajinikanth and K. Latha, Setpoint weight PID controller
15 tuning for unstable system using heuristic algorithm, Archives of
h1(cm)

Control Science, vol. 22, no.4, pp.481-505, 2012.

14 [2] N.Deepa,Design and Implementation of Neuro controllers for A

Two-Tank Interacting Level Process, M.E.Thesis,Annamalai
University,2014.
13
[3] George Stephanopoulos, Chemical Process Control-An
4000 4500 5000 5500 Introduction to Theory and Practice, PHI Learning Private
time(sec) Limited New Delhi, 2009.

Fig 20. Regulatory response of h1 with PI and PSO-PI due [4] Kennedy J, Eberhart R. Particle swarm optimization. In: Proc IEEE
to load variation in -10% from qL2. int conf neural networks, vol. IV, Perth, Australia; 1995. p.
194248.

[5] Eberhart RC, Shi Y. Particle swarm optimization: development,

700 PSO applications and resources. In: Proc congress on evol comput,
PI Seoul, Korea; 2001 . p. 816.

600
qin(cm /min)

[6] Ziegler.G and NicholsN.BOptimum Setting for

AutomaticControllers, Trans ASME.64pp. 759.2003
3

500 [7] Bhuvaneswari N S, Praveena R, Divya R, System Identification

and Modeling for Interacting and Non- Interacting Tank
System Using Intelligent Techniques, SV:1208.1103V,
August 2,2006.
400 [8] Donald R.Coughnowr, Process Systems Analysis and Control,
McGraw Hill International Editions, Second Edition, 1991.
300
4000 4500 5000
time(sec)
Fig 21. Regulatory response of qin with PI and PSO-PI
due to load variation in -10% from qL2.

TABLE IX
COMPARISON OF PERFORMANCE MEASURE OF
WITH PI AND PSO-PI FOR REGULATORY

Regulatory Response h2 (6-8)cm

Controller +10% from qL2 -10% from qL2
ts(sec) ISE ts(sec) ISE
PSO-PI 1036 2824 1045 2826
PI 1460 9321 1277 9311

VI. CONCLUSION
The non-linear mathematical model of a two-tank
interacting process was derived. To improve the performance
of closed loop control a PSO-PI was designed and
implemented for a two-tank interacting process. The servo
and regulatory responses were obtained with PSO-PI. The
performances of PSO-PI were compared with that of
conventional PI controller in simulation. The performance
measures were tabulated. It is observed that the PSO-PI gives
is better performance on terms of less integral square error,
faster settling time and without oscillation.

280 www.erpublication.org