Вы находитесь на странице: 1из 4

LEVEL CONTROL OF HORIZONTAL CYLINDRICAL TANK USING

INTELLIGENT CONTROL TECHNIQUES


1. M. Ulaganathan M.Tech., Assistant Professor (ulagu.er@gmail.com), 2. K. Abinaya, UG student
(abinaya.segar@gmail.com), 3. M.Aarthi, UG student (aarthi_m101@yahoo.co.in)
Department of Electronics and Instrumentation
Easwari Engineering College
ABSTRACT: This paper proposes control techniques for
the nonlinear level process of a horizontal cylindrical
tank. The control techniques used are Neuro controller
and Fuzzy Logic Controller which are implemented
using LabVIEW. Based on the performance criteria, it
could be concluded that Fuzzy Logic Controller
effectively improves the performance of the system.

KEYWORDS: Fuzzy Logic Controller, Horizontal


cylindrical tank, Neuro Controller, Nonlinear Level
Process
I.

INTRODUCTION
Horizontal cylindrical tank is unique in its shape,
as its area varies with increase in height. Many process industries use horizontal tank for storage of oil or
chemical liquid since it has a wide surface area. Level
control of tanks is necessary to maximize productivity
and efficiency and to ensure safety. Poor operations
include overfilling of the tanks due to which catastrophic reactions may take place.
Traditionally, a nonlinear process has to be
linearized first before an automatic controller can be
effectively applied. This is typically achieved by
adding a reverse nonlinear function to compensate for
the nonlinearity. There was a proposal in 2005 [1]
which showed the level control of a horizontal tank
implemented using Fuzzy-PID Cascade controller.
The application of a neural network trained for a
given set of inputs and output was published in 1993
[2] which was used for level control in petrochemical
tanks.
In this paper, the first section describes the
process and the identification of transfer function. In
the second section, we will discuss the conventional
control technique and its tuning method. The third
section illustrates the training of a neural network and
its implementation as a controller. Finally the Fuzzy
logic control (FLC) technique is discussed and the
most effective control method is established based on
the comparison of Integral Absolute error (IAE).

Fig.I Schematic diagram of the process

Diameter
Length
Height above the ground
Inflow

0.3m
0.45m
1.2m
120LPH

Table I: Dimensions of horizontal tank process

The process described in fig I shows that the


nonlinear tank level is controlled by final control
element fitted in the inflow side of the setup. The final
control element used in the process is a equal
percentage valve. The opening and closing percentage
depends on the current level of the fluid in the tank.
The dimensions of the tank are declared in table I. The
current level is measured by means of a differential
pressure transmitter (DPT). The output of DPT is
given to NIC-DAQ card which is interfaced with a
computer through serial communication. On receiving
the level value the LabVIEW program takes the
necessary control action. Through NIC-DAQ the value
is given to the final control element. To avoid damage
to the pump due to high pressure a by-pass line has
been provided.
B. MATHEMATICAL MODEL

II. PROCESS DESCRIPTION


A.SCHEMATIC DIAGRAM

Proceedings of the International Conference on Trends in Industrial Measurements and Automation (TIMA-2013),
Department of Instrumentation Engineering, MIT Campus, Anna University, Chennai-44.

Page 137

Fig II Side view of the cylindrical tank

The mathematical model is derived by


considering the side view of the tank of radius R and
level in the tank (H). The nonlinearity of the level is
calculated by subtracting the area of the triangle from
the area of the sector as shown in fig II. The resulting
area A is differentiated to obtain a term for changing
level in the tank.
2

r(R-H)

= ((2HR-H2)-1/2) [(

) + 6HR-2H2-2R2]

(1)

From (1), the Mass-Balance equation is derived as,


l*

= q1 - q2

Substituting for

from (1)

(2)

III. STEADY STATE CHARACTERISTICS


The mathematical model is designed using
LabVIEW based on (2) as derived above. We obtained
the steady state characteristics curve by varying the
inflow to the process in steps and plotting the graph
for inflow vs. steady state.

14-18

10.33

1.01

0.97

18-22

15.43

1.28

0.118

22-26

21.55

1.53

0.185

26-30

28.7

1.79

0.274

Table II : Transfer function by gain scheduling of the seven regions

A. SYSTEM IDENTIFICATION
System identification process is carried out using
the system identification function ident in MATLAB.
The input (inflow) and output (level) data from the
real time process in open loop arrangement applied to
the ident tool to obtain the transfer function of the
system. We obtained the transfer function for
maximum inflow as
We verified the real time transfer function to be
same as that of the simulation results. Hence the
estimated transfer function is used in the following
sections for designing the controllers.
IV. CONVENTIONAL CONTROLLER
We designed PI controller for each of the
seven regions defined in table II using gain
scheduling. The gain scheduling is achieved by tuning
the PI parameters (i.e. Proportional gain, integral time,
derivative time) using Ziegler Nichols tuning rules.
The controller is designed in such a way that
depending upon the input set value, the PI values for
the corresponding region is fired. This closed loop
system is implemented in LabVIEW and the response
curve is obtained.
A.

Fig IV Steady State Characteristics for inflow values ranging from


1-32 LPH

The Steady state characteristics graph shown in


fig IV is divided into seven regions. We obtained the
transfer function for every region and tabulated the
result as follows:
Rise
Steady
Transfer
time
Region
state
K
Function
(cm)
(cm)
0-6

1.15

0.25

0.196

6-10

3.19

0.51

0.176

10-14

6.25

0.77

0.138

ZIEGLER NICHOLS TUNING


We implemented the Ziegler Nichols tuning by
obtaining sustained oscillations using a P controller.
The PI parameters were calculated using the Ziegler
Nichols rules.
Region
(cm)
0-6
6-10
10-14
14-18
18-22
22-26
26-30

Proportional
gain (Kp)
58.55
25.62
13.64
7.16
7
9.123
11.54

Integral
gain (Ki)
0.038
0.028
0.031
0.026
0.033
0.032
0.028

Table III : PI parameters calculated for the seven regions of gain


scheduling using Ziegler Nichols tuning rules.

Proceedings of the International Conference on Trends in Industrial Measurements and Automation (TIMA-2013),
Department of Instrumentation Engineering, MIT Campus, Anna University, Chennai-44.

Page 138

Fuzzy toolbox in LabVIEW as shown in fig IX. The


Mamdani-type Fuzzy Inference System (FIS) is then
used to map the inputs and output based on the fuzzy
logic.

Fig V: Closed loop response by Ziegler Nichols tuning for


consecutive step inputs of 12cm, 20cm, 15cm

The PI controller designed for all the regions as


defined in table III is used to obtain the closed loop
response shown in fig V.
IV.NEURAL NETWORK
We observed inaccurate control and large
overshoot in conventional controllers which could
cause instability for certain set values . To overcome
these drawbacks, we opted for Neuro Controller.
To implement Neuro Controller, we obtained real
time data by running the open loop system and trained
the neural network with the inputs (set point and error)
and target output (percentage valve opening). We
designed the neural network in MATLAB using
nntool function and simulation block using Gensim
function. The Neuro controller designed in MATLAB
is called in LabVIEW using mathscript function.

Fig IX: Membership functions for inputs(error and change in error)


and output(percentage valve opening) for FLC defined in LabVIEW

Fig X : Closed loop response curve using FLC for step input set
values of 12cm, 20cm, 15cm

VI. REALTIME IMPLEMENTATION

Fig VIII : Closed loop control using Neuro Controller for a set value
of 12cm, 20cm, 15cm

The controller called in LabVIEW is then


interfaced with the process to obtain the the closed
loop response shown in fig VIII.
V. FUZZY LOGIC CONTROLLER
We observed slow response in the neural
controller which we expected to overcome using FLC.
To design a fuzzy logic controller, we first
declared the universe of discourse for the input
variables (error and change in error) and the output
variable (percentage of valve opening). We defined the
membership functions and the rule base using the

Fig XI: Real time Hardware setup of the nonlinear horizontal tank
system

We implemented the Closed loop control using


PID controller and Fuzzy logic controller in real time
by interfacing the hardware of the process with
LabVIEW environment in a computer using NI CDAQ cards as shown in fig XI. Since the system
transfer function is proved to be equal in simulation

Proceedings of the International Conference on Trends in Industrial Measurements and Automation (TIMA-2013),
Department of Instrumentation Engineering, MIT Campus, Anna University, Chennai-44.

Page 139

and real time, we used the same controllers designed


in simulation for the real time control.

The response of the controllers is observed and


compared as shown in fig XIV and the performance is
evaluated using IAE.

Zeigler
Nichols
Neural
Network
Fuzzy Logic
Controller
Fig XII : Closed loop control of the real time process using PI
controller for a set value of 12cm

Rise

Settling

IAE

time

time

240

1450

4458.7

120

1250

3768.4

100

900

3557.8

Table IV : Comparison of the controller action based on the


performance criteria rise time, settling time, IAE

The Conventional controller showed good


response time and precision but had large overshoots.
Also, due to linearization of the non-linear process,
several PI controllers were used.
The Neural controller overcomes the above
drawbacks but the response had longer settling times.
The Fuzzy controller rules out linearization. It
has good time domain characteristics. The
experimental results show that the response has lesser
oscillations and absence of overshoot.
VIII. CONCLUSION:
We concluded that the Fuzzy Logic Controller
outperforms both the conventional PID and Neural
controller, in the level control of a nonlinear horizontal
cylindrical tank system.

Fig XIII : Closed loop control of the real time process using FLC for
step input set values of 12cm, 20cm, 15cm

VII. RESULT:
The response of the closed loop system was
studied and the drawback of conventional controller
and need for intelligent control is observed. The
performance of the controllers is compared using IAE.

REFERENCE
[1] Satean Tunyasrirut, and Santi Wangnipparnt Level
Control in Horizontal Tank by Fuzzy-PID Cascade
Controller - World Academy of Science, Engineering
and Technology 25 -2007.
[2] Tani. T, Umano. M, Sato. T, Tanaka. K,
"Application of neural network to tank level control of
petrochemical plants," Industrial Electronics, Control,
and Instrumentation, 1993. pp.321,326 vol.1, 15-19
Nov 1993
[3] J. G. Ziegler and N. B. Nichols, Optimum settings
for automatic controller, ASME Trans. Vol64, 1942,
pp.759-768 And Automatic setting for Automatic
Controller
[4] W. K. Ho, C. C. Hang, and J. H. Zhou,
Performance and gain and phase margins of wellknown PI Tuning formulas, accepted for publication
in IEEE Trans. Contr. Syst. Technol., 1995.
[5] Stephanopoulos G (1990) Chemical Process
Control, Prentice Hall, Englewood Cliffs, New Jersey,
chapter 16

Fig XIV: Comparison of the response of conventional controller,


neural controller and fuzzy logic controller for consecutive set
values of 12cm, 20cm, 15cm.

Proceedings of the International Conference on Trends in Industrial Measurements and Automation (TIMA-2013),
Department of Instrumentation Engineering, MIT Campus, Anna University, Chennai-44.

Page 140

Вам также может понравиться