See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/313860636

Design of Static and Dynamic Decoupler for

Coupled Quadruple Tank Level System

Article · February 2017

CITATIONS READS

0 69

2 authors:

Hasan Vhora Jignesh Patel

Nirma University Nirma University
2 PUBLICATIONS 2 CITATIONS 23 PUBLICATIONS 25 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

M.tech Project on Design and Development of single tank level system View project

Design and Investigation of Advanced Control Schemes for Quadruple Conical Tank System View project

All content following this page was uploaded by Jignesh Patel on 21 February 2017.

The user has requested enhancement of the downloaded file.

NIRMA UNIVERISTY JOURNAL OF ENGINEERING AND TECHNOLOGY VOL. 5, N0. 2, JULY-DEC 2016 8

Design of Static and Dynamic Decoupler for

Coupled Quadruple Tank Level System
Hasan Vhora and Jignesh Patel

Abstract—Multi Inputs Multi Output (MIMO) processes that can be used to do such performance analysis on
are very common in industrial environment. These loops laboratory scale [4]. Fig. 1. shows a schematic diagram
comprise multiple measurement (sensor) and manipulating of quadruple tank system it is a simple system with four
(control) signals. The quadruple tank system is popular test
bench setup to demonstrate various properties of MIMO interconnected tanks provided with the inflows by two
process. Major challenge faced while designing a control pumps. Inflows are inputs and height of water in lower
system for MIMO processes is the interaction (or coupling) two tanks are considered as outputs. This will make the
between inputs and outputs. In such cases conventional system as multivariable system with two inputs and two
multi loop PID control works when the coupling is less or outputs.
weak. When interactions increases then the conventional
multi loop PID control fails. Hence, need arises to provide
control solutions for highly coupled system and one of
the most primitive approach is to implement decoupler
control. Thus, attempt is made to understand and realize
the problem for controlling a highly coupled process with
help of quadruple tank system. Design of decoupler using
LabVIEW and its validation are the prime work of the
presented paper.
Index Terms—Decoupler Design, LabVIEW, MIMO
Coupled Process, PID Control, Quadruple Tank System

I. I NTRODUCTION
.

M AJORITY of industrial processes comprises of

multiple manipulated variable and multiple con-
trol variable. There are more than one measurement and
control signals to look out for [1]. Major challenge in this
case for a control engineer is that such processes cannot
be controlled by single loop PID control [2]. Hence,
advance process control schemes involving multiple PID
controllers are to be considered [3]. Major problem in
control of multivariable system is interaction or coupling
effect present between the inputs and outputs of the sys-
tem. In such cases disturbance acting on one control loop
affects whole system, this phenomenon approves that
Fig. 1: Schematic diagram of quadruple tank system
there is presence of coupling between input and output of
the system. There exists only a handful of multivariable This test process is quite popular as benchmark MIMO
laboratory processes to illustrate the effect of coupling process [5]. It is well establish platform to show traits
between input and output on closed-loop control perfor- of a multivariable system [6, 7]. Advantage of using
mance. Quadruple tank system is one of such process quadruple tank system as a test bench for multivariable
Hasan Vhora is post graduate student at Instrumentation and Con- system is that one can change the coupling between
trol Engineering Section , Institute of Technology, Nirma University, inputs and output of the system by adjusting the diverter
Ahmedabad, Gujarat, India, Email: 14micc30@nirmauni.ac.in. valves D1 and D2 [8].
Jignesh Patel is with Instrumentation and Control Engineering,
Institute of Technology, Nirma University, Ahmedabad, Gujarat, India, Common challenge while controlling quadruple tank
Email: jbpatel@nirmauni.ac.in. system is that when setup is taken under consideration
NIRMA UNIVERISTY JOURNAL OF ENGINEERING AND TECHNOLOGY VOL. 5, N0. 2, JULY-DEC 2016
Fig. 2: Block diagram of close loop multivariable system with decoupler control
keeping inputs outputs strongly coupled classical PID
control fails as control action resulted from h1 will
manipulate pump 2 rather than pump 1 and control
action resulted from h2 will manipulate pump 1 rather
than pump 2 making system un controllable [9]. Hence,
need arises of applying advance control algorithms like
decoupler control in order control such processes as the
decouplers have been proven successful to remove the
coupling effect between input and output [10].
Fig. 2. shows block diagram of multi PID control
with decoupler. In this scheme the important task is
to design decouplers. The decoupler, T12 and T21 are
lead-lag type compensators by its structure. T21 is
designed so as to cancel Y21 which arises from the
undesirable process interaction betweenU1 & h2 .
This cancellation will occur at the Y2 summer if
decoupler output U2 satisfies
Gp11 U11 + Gp22 U21 = 0
Substituting Up21 = T21 U11 and factoring gives,
(Gp21 + Gp22 T21 )U11 = 0
Note that U11 (S) is not zero because U11 is controller
output and is time dependent,
T21 = Gp21 /Gp22
Similarly a design for T12 can be derived by imposing
the requirement that U22 has no net effect on Y1 . Thus,
the compensating signal U12 and the process interaction
9
due to Gp12 should cancel at the Y1 summer.
(Gp12 + Gp11 T12 )U22 = 0
Hence, ideal decoupler is given by,
T12 = Gp12 /Gp11
These ideal decouplers acts very similar to ideal
feed forward controllers. These decoupler designed
are nothing but lead-lag compensator. There are two
approach to implement decouplers
i. Static Decoupler
ii. Dynamic Decoupler
Static decoupler uses only dc gain part of process
transfer function, while, dynamic decoupler uses transfer
function along with its dynamic part. Ideally dynamic de-
coupler give better response as it comprises dynamic part
compensation while static decoupler only compensate
the steady state part An effort is made to design static
and dynamic decoupler for a high coupled quadruple
tank system [11].
The next section provides the details on development of
model of quadruple-tank level system. Following that the
implementation of static and dynamic decoupler on Lab-
VIEW platform is presented. The fourth section of the
paper presents the control performance of the quadruple
tank level system for conventional PID control, static
decoupler control and dynamic decoupler. Comparative
results and discussion are presented in the last section.
NIRMA UNIVERISTY JOURNAL OF ENGINEERING AND TECHNOLOGY VOL. 5, N0. 2, JULY-DEC 2016 10

0.003448s+0.001157
II. E MPIRICAL MODELLING T21 = Gp12 /Gp11 = 0.001281s+3.613∗10−05

And, Static decouplers are

T12 = 10.6074
T21 = 64.7657
III. I MPLEMENTATION OF DECOUPLER
The implementation of the decoupler is performed
on LabVIEW [14] platform. To implement static type
of decoupler it is required to multiply dc-gain value to
the PID controller output so a simple arrangement with
multiplication block.

Fig. 3: . Open loop responses of level of both the tanks

In order to design decoupler it is required to have

transfer function of actual physical system. Hence, the
model of actual physical system is required to be ob-
tained [12]. In the present work modelling is done with
help of System Identification Tool of MATLAB [13].
In order to obtain process transfer functions, Gp11 and
Gp21 open loop response of system is captured keeping
supply to pump 2 (Input 2) zero and giving step change
to supply of pump 1 (input 1), h1 and h2 is recorded. Fig. 4: Static decoupler implementation using LabVIEW
Similarly to obtain Gp12 and Gp22 open loop response
of system is taken keeping supply to pump 1 (Input 2)
as zero and step change is given in supply of pump 2
(input 1), h1 and h2 is recorded. Open loop responses
obtained by this procedure is as shown in Fig. 3.
The obtained data is scaled in values of 0% to 100%
and then fed to MATLAB in order to find the transfer
functions. These obtained transfer functions are then
reduced into first order transfer functions for simplicity.
The simplified first order transfer functions considered
in the decoupler design are as follows:
0.001281
Gp11 = s+0.0244
0.001571
Gp12 = s+0.00282
0.00343
Gp21 = s+0.00871
0.009546
Gp22 = s+0.3355 Fig. 5: Dynamic decoupler implementation using Lab-
VIEW
Dynamic decouplers are
Fig. 4. shows the block diagram implemented in
0.001571s+3.883∗10−05
T12 = Gp21 /Gp22 = 0.001281s+3.613∗10−06
LabVIEW to realize static decoupler. Here controller
NIRMA UNIVERISTY JOURNAL OF ENGINEERING AND TECHNOLOGY VOL. 5, N0. 2, JULY-DEC 2016
Fig. 6: Response of strongly coupled quadruple tank with (a) Conventional PID control, (b) Static decoupler control,
(c) Dynamic decoupler control
outputs are multiplied by respective decoupler and then
are added to simultaneous controller outputs [15] their
sum is realized and control action is h2 and h1 in form
of voltage supply to pump 2 and pump 1 respectively by
obtaining its value in real-time after proper conditioning
by using appropriate hardware peripheral like NI DAQ
6009 [16]. transfer functions are being written and
supplied by PID controllers outputs.
Fig. 5. shows the block diagram implemented in Lab-
VIEW to realize dynamic decouplers where control and
simulation toolbox [17] is employed in order to realize
decoupler transfer functions in real-time. Dynamic de-
couplers requires transfer functions to be realized hence
control and simulation toolbox is employed where Here
controller outputs are provided to dynamic decouplers
transfer functions and then are added to simultaneous
controller outputs their sum is realized and control action
is h2 and h1 in form of voltage supply to pump 2 and
pump 1 respectively by obtaining its value in real-time
after proper conditioning by using appropriate hardware
peripheral
IV. R ESULTS AND DISSCUSSIONS
Fig. 6. (a) shows responses of quadruple tank sys-
tem by implementing Conventional PID control with
Kp1 = 5, Ki1 = 0.001 Kd1 = 0.100 & Kp2 = 5,
Ki2 = 0.001, Kd2 = 0.100, it can be seen that h1
tends to zero while h2 tends to uncontrolled values
implying that tank 1 runs dry while tank 2 overflows
making it an completely unstable system, This shows
primary limitation of conventional multiple PID loops
in control of MIMO process with strong coupling. Fig.
6. (b) shows responses of quadruple tank system by im-
plementing static decoupler control this response much
stable than that of conventional PID control and dynamic
decoupler control implying using LabVIEW it is most
11
appropriate technique to control aforementioned system.
Fig. 6. (c) shows responses of quadruple tank system by
implementing dynamic decoupler control this response
much stable than that of conventional PID control but
Static decoupler shows better transient response and less
oscillations at steady state implying that this control
strategy works but its not the most appropriate technique
to control aforementioned system. The dynamic decou-
pler implementation needs accurate modeling. It seems
that in the present work the accuracy of the model cannot
achieved. In case of model mismatch situation, the static
decoupler is more appropriate design.
V. C ONCLUSION
It clearly seen that conventional PID control fails
with high coupled system. Responses of static and dy-
namic decoupler controlled systems are relatively stable
and oscillate within a small region which is makes
it marginally stable. Ideally dynamic decoupler shall
provide more stable response than static decoupler. But
in this case response of static decoupler shows better
response than dynamic decoupler. Possible reason for
this contradiction to theory is inaccurate modeling of
the process. The accuracy of model is very vital for
the successful implementation of dynamic decoupler. In
the present setup the model mismatch is very much
expected, hence, the dynamic decoupler shows a re-
sponse with relatively more oscillations and overshoot
than that of static decoupler based design. Another
observation is that the control algorithm is implemented
on windows based LabVIEW platform. It has been
observed that the control action takes some finite time
in execution. Such situation leads to time delay in the
control loop and contributes to instability of the system.
Dedicated real time control platform can surely improve
the control performance. Dynamic and static decouplers
 NIRMA UNIVERISTY JOURNAL OF ENGINEERING AND TECHNOLOGY VOL. 5, N0. 2, JULY-DEC 2016 12

have been successfully designed for strongly coupled Hasan Vhora received the B.Tech. degree
MIMO system. The decoupler control scheme has been in Instrumentation and Control Engineering
from Dharmsinh Desai University, Nadiad,
implemented using LabVIEW platform. The controlled India, in 2014 and is currently pursuing
variable response shows the improved performance in M.Tech. degree in Instrumentation and Con-
comparison to conventional PID controller based system. trol Engineering (Control and Automation)
from Nirma University, Ahmedabad, India.
His area of interests are Control Engineering
ACKNOWLEDGMENT and Process Control.

The authors would like to extend their heartfelt grat-

itude to Instrumentation & Control Engineering Sec-
Dr. Jignesh Patel is Sr. Associate Pro-
tion, Department of Electrical Engineering, Institute of fessor, Instrumentation and Control Engi-
Technology, Nirma University, Ahmedabad for providing neering, Electrical Engineering Department
timely needed resources and assistance. at Institute of Technology Nirma Univer-
sity, Ahmedabad, India. He has total 20
years of teaching and research experience in
R EFERENCES academia. He received his BE in Instrumen-
tation and Control Engineering from Gujarat
[1] Thomas E. Merlin, “Process Control Designing Processes and University, India; M.Tech. in Instrumentation
Control Systems for Dynamic Performance”, McGraw-Hill Inc., Engineering from IIT Kharagpur, India; and
1995. PhD in Instrumentation and Control Engi-
[2] C.G. Goodwin, S.F. Graebe, M.E. Salgado, “Control System neering from Kadi Sarv Vishwavidyalay University, Gandhinagar,
Design”, Prentice-Hall,2001 India. He was conferred with Institute Silver Medal-2002 at Indian
[3] Dale E. Seaborg, Thomas F. Edgar , Duncan A. Mellichamp, Institute of Technology, Kharagpur.
“Process Dynamics and Control”, John Wiley Sons Inc., 2004..
[4] Karl Henrik Johansson, Alexander Horchz, Olle Wijkz, Anders
Hanssonz, “Teaching Multivariable Control Using the Quadruple-
Tank Process”, IEEE Conference on Decision and Control,
Phoenix, AZ, 1999.
[5] Alvarado, D. Limon, W. Garca-Gabn , T. Alamo , E.F. Camacho,
“An Educational Plant Based on the Quadruple Tank Process”,
International Federation of Automatic Control, Madriad , Spain,
2006.
[6] Jignesh Patel and Hasan Vhora, “Development of Multi Input
Multi Output Coupled Process Control Laboratory Test Setup ”.
International Journal of Advanced Research in Engineering and
Technology, Volume 7 Issue 1 , February 2016, pp. 97-104.
[7] Effendi Rusli, Siong Ang. , Richard D. Braatz, “A Quadruple-
Tank Process Control Experiment”, University of Illinois at
Urbana Champaign, 2004.
[8] Karl Henrik Johansson, Alexander Horchz, Olle Wijkz, Anders
Hanssonz, “Teaching Multivariable Control Using the Quadruple-
Tank Process”, IEEE Conference on Decision and Control,
Phoenix, AZ, 1999.
[9] Vzquez, F.; Morilla, F.; Dormido, S., “An iterative method for
tuning decentralized PID controllers”, Proceeding of the 14th
IFAC World Congress, pp. 491-496, 1991.
[10] K.J. strm, K.H. Johansson, Q.-G. Wang., “Design of decoupled PI
controllers”, IEE Proceedings: Control Theory and Applications,
2002.
[11] Grebeck, Micheal, “A Comparision of Controllers for Quadruple
Tank System”, Internal Report, Dept. of Automatic Control, Lund
University, Lund, Sweden, 1998.
[12] Tomi Roinila, Matti Vilkko, Antti Jaatinen, “Corrected Mathe-
matical Model of Quadruple Tank Process”, The International
Federation of Automatic Control, Seoul, Korea, 2008.
[13] MathWorks, “System Identification Toolbox, ”, Available:
http://http://in.mathworks.com/products/sysid/.
[14] National Instruments, “NI LabVIEW PID and Fuzzy Logic
Toolkit for Windows”, National Instruments, Available:
http://sine.ni.com/nips/cds/view/p/lang/en/nid/209054.
[15] National instruments, “PID Control”, Instruments, Available:
http://www.ni.com/white-paper/6440/en/.
[16] National Instruments, “USER GUIDE NI USB-6008/6009”,
Available: http://www.ni.com/pdf/manuals/371303n.pdf.
[17] National instruments, “Control Design& Simulation Toolbox ”,
Instruments, Available: http://www.ni.com/labview/cd-sim/.

View publication stats