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

net/publication/301619534

Optimized PID Controller for Magnetic Levitation System

Conference Paper  in  IFAC-PapersOnLine · February 2016

DOI: 10.1016/j.ifacol.2016.03.151

CITATIONS READS

45 1,941

3 authors, including:

Shekhar Yadav Santosh kumar Verma

Madan Mohan Malaviya University of Technology, Gorakhpur (U.P.) India BIT Sindri, Dhanbad, Jharkhand, India
28 PUBLICATIONS   221 CITATIONS    11 PUBLICATIONS   191 CITATIONS   

SEE PROFILE SEE PROFILE

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

fractional order pi controller design for non-monotonic phase system View project

Modeling and Reduction of large scale system View project

All content following this page was uploaded by Shekhar Yadav on 10 January 2019.

The user has requested enhancement of the downloaded file.

4th International Conference on Advances in Control and
4th International
Optimization Conference
of Dynamical on Advances in Control and
Systems
4th International Conference on Advances in Control and
4th International
Optimization
February
4th 1-5, of Conference
Dynamical
2016.
International NIT on Advances
Advances
Systems
Tiruchirappalli,
Conference on in Control
Available
India
in Control and
onlineand
at www.sciencedirect.com
Optimization of Dynamical Systems
Optimization
February 1-5, of
Optimization Dynamical
2016. Systems
NIT Tiruchirappalli, India
February 1-5, of Dynamical
2016. Systems
NIT Tiruchirappalli, India
February
February 1-5,
1-5, 2016.
2016. NIT
NIT Tiruchirappalli,
Tiruchirappalli, India
India
ScienceDirect
IFAC-PapersOnLine 49-1 (2016) 778–782
Optimized PID Controller for Magnetic Levitation System
Optimized
Optimized PID
PID Controller
Controller for
for Magnetic
Magnetic Levitation
Levitation System
System
Optimized
Optimized PID
PID Controller
Controller *
for
for Magnetic
Magnetic
**
Levitation
Levitation
***
System
System
Shekhar Yadav , S. K. Verma , S. K. Nagar
* ** ***
Shekhar
Shekhar Yadav
Yadav *, S. K. Verma**, S. K. Nagar***
*, S. K. Verma**, S. K. Nagar***
Shekhar Yadav *, S. K. Verma**, S. K. Nagar***
Shekhar Yadav , S. K. Verma , S. K. Nagar
Department of Electrical Engineering, Indian Institute of Technology (BHU),
Department
Department of
of Electrical Engineering,
ElectricalVaranasi-221005,
Engineering, Indian
Indian
U.P.,Institute
India of
Institute of Technology
Technology (BHU),
(BHU),
Department
Department of of Electrical Engineering,
Varanasi-221005,
Electrical Engineering, Indian
U.P.,
Indian Institute
India
Institute of
of Technology
Technology (BHU),
(BHU),
Varanasi-221005,
e-mail: U.P., India
syadav.rs.eee@iitbhu.ac.in*,
Varanasi-221005, U.P., India
India
e-mail:
e-mail: syadav.rs.eee@iitbhu.ac.in*,
Varanasi-221005, U.P.,
syadav.rs.eee@iitbhu.ac.in*,
santosh.rs.eee13@iitbhu.ac.in**, sknagar.eee@iitbhu.ac.in***
e-mail: syadav.rs.eee@iitbhu.ac.in*,
syadav.rs.eee@iitbhu.ac.in*,
santosh.rs.eee13@iitbhu.ac.in**,
e-mail: sknagar.eee@iitbhu.ac.in***
santosh.rs.eee13@iitbhu.ac.in**,
santosh.rs.eee13@iitbhu.ac.in**, sknagar.eee@iitbhu.ac.in***
sknagar.eee@iitbhu.ac.in***
santosh.rs.eee13@iitbhu.ac.in**, sknagar.eee@iitbhu.ac.in***
Abstract: In this paper, an optimized proportional-integral-derivative (PID) controller is designed to control the ball
position ofIn
Abstract:
Abstract: thethis
In this paper,
paper, an
magnetic optimized
levitation
an optimized proportional-integral-derivative
system (MLS). The electromagnetic(PID)
proportional-integral-derivative force controller
(PID) of the MLS
controller is
is designed to
to control
is controlled
designed the
the ball
by sensing
control the
ball
Abstract:
position
Abstract: of In
the
In this
this paper,
magnetic
paper, an
an optimized
levitation
optimized proportional-integral-derivative
system (MLS). The electromagnetic
proportional-integral-derivative (PID)
force
(PID) controller
of the MLS
controller is
is designed
is to
controlled
designed to control
by ofthe
timeball
sensing
control the the
ball
position of the ball with the
magnetic help of system
levitation the infra-red
(MLS). (IR) sensors.
The The systemforce
electromagnetic performance
of the MLS is improved in terms
is controlled by sensing &
the
position of the magnetic
the ball with levitation
the help system (MLS). ofThe electromagnetic force of
of the MLS is
is controlled by
by sensing the
position
frequency
position ofdomain
of the magnetic
ball by
with help of
levitation
optimizing
the ofthethe
systeminfra-red
(MLS).
theparameters
infra-red (IR)
(IR) sensors.
The
the The
The system
electromagnetic
PID controller
sensors. performance
force
using
system greythewolf
performance MLS is improved
isoptimizer
improved in
in terms
controlled
(GWO).
terms of
of time
sensing
The GWO
time &
the
&
position
frequency oftunes
the ball
domainball with
theby the help
helpofof
optimizing ofthetheparameters
infra-red (IR)
of sensors. The system
system performance is improved
improved in terms
terms ofsuch
time &
position
algorithmof
frequency the
domain with
by the
parameter
optimizing the
the
the infra-red
PID of the
(IR)
controller
parameters PID
sensors.
while
the controller
The
PIDminimising
controller using
the
using grey
grey wolf
performancewolfisoptimizer
performance index of (GWO).
optimizer in
the
(GWO). The
systemof
The GWO
time
GWO &
as
frequency
algorithm
frequency
integral domain
tunes
domain
time theby
by
weighted optimizing
parameter
optimizing
absolute of the
the
the
error parameters
PID
parameters
(ITAE) and of
controller
of the
the PID
while
PID
integral controller
minimising
controller
time weighted using
the
using grey
grey
square wolf
performance
wolf
error optimizer
index of
optimizer
(ITSE). The (GWO).
the
(GWO). The
system
The
effectiveness GWO
such
GWO
of as
the
algorithm tunes the parameter of the PID controller while minimising the performance index of the system such as
algorithm
integral
proposedtimetunes the
the isparameter
weighted absolute of the
error PID
PID controller
(ITAE) and while
integral minimising
time tuningthe
weighted performance
square error index of the
the system
system such as
algorithm
integral tunes
controller
time weighted parameter
absoluteof
validated by the
errorcomparing
(ITAE)controller
it with
and while
the
integral minimising
classical
time weighted the error (ITSE).
performance
criterion.
square indexThe
(ITSE). of effectiveness
The effectiveness of
of the
such as
the
integral
proposed
integral time
time weighted
controller
weightedis absolute
validated
absolute error
by
error (ITAE)
comparing
(ITAE) and
it
and integral
with the
integral time weighted
classical
time tuning
weighted square error
criterion.
square error (ITSE).
(ITSE). The
The effectiveness
effectiveness of
of the
the
proposed controller is validated by comparing it with the classical tuning criterion.
© 2016, IFAC
Keywords:
proposed
proposed
Keywords:
Magnetic
controller
levitation Federation
(International
controller is
is validated
system,
validated by
PIDofcontroller,
by comparing
comparing Automatic
it
it with
Grey
with the
wolf optimizer.
Control)
the classical
classical Hosting
tuning
tuning by Elsevier Ltd. All rights reserved.
criterion.
criterion.
Keywords: Magnetic
Magnetic levitation
levitation system,
system, PID PID controller,
controller, Grey
Grey wolf
wolf optimizer.
optimizer.
Keywords: Magnetic levitation
Keywords: Magnetic levitation system,
system, PID PID controller,
controller, Grey
Grey wolf
wolf optimizer.
optimizer.
transient and steady-state performance. The parameters of
1. INTRODUCTION transient and steady-state
steady-state performance. Thetuning
parameters of
1. INTRODUCTION transient
transient and
PID controller
and are tuned performance.
steady-state using classical
performance. The
The parameters
criterion
parameters of
of
1.
1. INTRODUCTION
INTRODUCTION PID controller
transient and are
steady-state tuned using
performance.classical The tuning criterion
parameters of
Magnetic levitation 1. systems
INTRODUCTION have been widely used in PID
such
PID controller
controller are
as Ziegler-Nichols
are tuned
tuned using
(Z-N)
using classical
tuning rules.tuning
classical The Z-N
tuning criterion
rules
criterion
Magnetic
Magnetic
magnetic levitation
levitation
bearings, systems
systems
high-speed have
trains,been
have been
vibration widely
widely used
used of
isolation in
in such
PID
such
give as
as Ziegler-Nichols
controller are
an Ziegler-Nichols
educated guess(Z-N) tuned (Z-N)
using tuning
for tuning rules.
classical
rules.
the parameter The
tuning
The Z-N
Z-N
values rules
criterion
rules
and
Magnetic
magnetic
Magnetic levitation
bearings,
levitation systems
high-speed
systems have
trains,
have been
vibration
been widely
widely used
isolation
used in
of
in such
such as
give an Ziegler-Nichols
as educated
Ziegler-Nichols (Z-N)
guessfor for
(Z-N) tuning
the
tuning rules.
parameter
rules. The
Thethan Z-N
values rules
Z-Ngivingand
rules
magnetic
sensitive bearings,
machinery, high-speed
levitation trains,
of wind vibration
tunnel isolation
models, of
etc., give
provide
give an an a educated
starting
an a educated
educated guess
point
guessforforfor
fine
for the
the parameter
tuning, rather
parameter values
valuesgivingand
and
magnetic
sensitive
magnetic bearings,
machinery,
bearings, high-speed
levitation
high-speed trains,
of wind
trains, vibration
tunnel isolation
models, of
etc., provide
give starting point
guess fine tuning,
the rather
parameter than
values and
sensitive
C. R. Knospe
sensitive machinery,
and E.
machinery, levitation
G. Collins
levitation of
of windvibration
(1996).
wind tunnel
Several
tunnel
isolation
models,
models,
of
etc.,
magnetic
etc.,
provide
the
provide finalaasettings
starting
starting pointspoint
point for for fine
for fine tuning,
the PID
tuning, rather
rather than
controller, than giving
Katsuhiko
giving
C. R.
sensitive Knospe and
machinery, E. G. Collins
levitation of (1996).
wind Several
tunnel magnetic
models, etc., the
provide
the final
final a settings
starting
settings points
To pointspoint for
for
for the
fine
the PID
tuning,
PID controller,
rather Katsuhiko
than giving
C. R.
suspension
C.
suspension
C. R.
Knospe
R. Knospe
Knospe and
systems
and E.
systems
and
E. G.
have
E.have Collins
been
G. Collins
G. Collins
beensystems
(1996).
developed
(1996). Several
developed
(1996).
Several
and
Several
and applied
magnetic
applied
magnetic
applied
magnetic
for
for
Ogata
the
Ogata
the
(2009).
final settings
(2009).
final
improve
points
To points
settings improve forthe
forthethe PID controller,
thetransient
transient
PID
performance
controller,
performance
controller,
Katsuhiko
of the
Katsuhiko
of the
Katsuhiko the
suspension
magnetically systems
levitated have been
transit developed and
in recent years, for
has Ogata
system,
Ogata (2009).
these
(2009). To
tuned
To improve
parameters
improve the
the transient
are
transient performance
optimized.
performance of
of the
suspension
magnetically
suspension systems
levitated
systems have
have been
transit
been developed
systems
developed in and
recent
and applied
years,
applied for
has
for system,
Ogata these
(2009). tuned
To parameters
improve the are optimized.
transient performance of the
magnetically
been investigatedlevitated
by Ruletransit R. et systems
al. (1980) in recent
and Ion years,
B ethas al. system,
system, these
these tuned
tuned parameters
parameters are
are optimized.
optimized.
magnetically
been investigated
magnetically levitated
by
levitated transit
Rule
transit R. systems
et al.
systems (1980)in
in recent
and
recent Ionyears,
B
years, et has
al.
has S. Mirjalili,
system, these S.tuned
M. Mirjalili,
parameters andare A. Lewis
optimized.(2014) introduces
been
(1988).
been investigated
Maglev
investigated by
systems
by Rule
Rule are R.
R. et
et al.
inherently
al. (1980)
(1980) and
nonlinear,
and Ion
Ion B
B et
unstable
et al.
al. S. Mirjalili, S. M.
M. Mirjalili,
Mirjalili, and A. A. Lewis
Lewis (2014) knownintroduces
(1988).
been Maglev
investigated systems
by Rule are R. inherently
et al. (1980) nonlinear,
and Ion unstable
B et al. S. Mirjalili,
aS. new
Mirjalili, S.
meta-heuristic
S. M.
M. Mirjalili,
Mirjalili, and
evolutionary
and A. A. Lewis
Lewis (2014)
algorithm
(2014) knownintroduces
introduces as
(1988).
(1988). Maglev
and are described
Maglev systems
by highly
systems are
are inherently
nonlinear
inherently nonlinear,
differential
nonlinear, unstable
equations
unstable aaGrey
S. new meta-heuristic
Mirjalili, S. evolutionary
and algorithm
(2014) introduces as
and are
(1988). described
Maglev by
systemshighly are nonlinear
inherently differential
nonlinear, equations
unstable new
aaGrey
new meta-heuristic
wolf optimizer
meta-heuristic evolutionary
(GWO)
evolutionary is algorithm
used to
algorithm known
optimize
known as
the
as
and
which
and are
are described
present
described by
by highlydifficulties
additional nonlinear differential
in controlling equations
these new wolf optimizer
meta-heuristic (GWO)
evolutionary is used to
algorithm optimize
known the
as
which
and
which
systems. present
arepresent
described
So the by highly
additional
highly
additional
design of
nonlinear
difficulties
nonlinear
difficulties
feedback
differential
in controlling
in
controller controlling
differential for
equations
these
equations
these
regulating
Grey Grey wolf
parameters
wolf
parameters
optimizer
of the PID(GWO)
optimizer
of the PID
controller.
(GWO)
controller.
is
is used
The GWO
used
The
to
to
GWO
optimize
algorithm
optimize
algorithm
the
the
which
systems. present
So the additional
design of difficulties
feedback in
controller controlling
for these
regulating Grey
parameters
searches wolfforofoptimizer
thethe PID
optimum (GWO)
controller.is
solution Theused
in the to
GWO optimize
desired algorithm the
range
which
systems.
the present
position So the additional
design
of the of
levitated difficulties
feedback in
controller
object controller controlling
is always for for these
regulating
a challenging parameters
searches forof thethe PID
optimum controller.
solution The
in GWO
the desired algorithm
range
systems.
the position So the
the
of design
the of feedback
levitated feedbackobject is always a regulating
challenging parameters
searches
space by for ofthe
updating theoptimum
PID
the controller.
position solution
of the The
in
wolves GWO
the desired algorithm
depending rangeon
systems.
the
taskposition
is So of
presented design
theby of
levitated
Vinodh object
Kumar controller
is always
et al. for
a
(2013). regulating
challenging
In most searches for the optimum solution in the desired range
the position
position of the
thebylevitated
levitated object is etalways
always space
searchesby updating
for the the
optimum position of
solution the wolves
in the depending
desired rangeon
task
the
task
cases isthe
is presented
of
presented
control by
systemVinodh
Vinodhand Kumaris
object
Kumar
energy et
supplyal. aa challenging
al. (2013).
(2013).challenging
requirementsIn most
In mostto
spacethe
space
the
by
by updating
location of
updating
location of
the
the
the
the position
prey, the
position
prey, the
of
of the
concept
the
concept
wolves
given
wolves
given
depending
by Seyedali
depending
by Seyedali
on
on
task
cases is presented
the control by
systemVinodhand Kumar
energy et
supplyal. (2013).
requirementsIn mostto the space by
Mirjalili updating
location of
(2015). the the
The position
prey, GWOthe of the
concept wolves
algorithm given depending
is by Seyedali
capable on
of
task
cases is
levitate presented
the control
the object by
systemVinodh
have and
a higher Kumar
energy level et
supplyal.
of (2013).
requirements
complexity. In most
Theto the location
Mirjalili of
(2015). the The prey, GWOthe concept
algorithm given is by Seyedali
capable of
cases
levitate the control
the object system
have and
a higherenergy levelsupplyof requirements
complexity. Theto the
Mirjalililocation
maintaining of
(2015).
the the The
balance prey, GWOthe
between concept
algorithm
the given is
exploration by Seyedali
capable
and of
the
cases
levitate
nonlinearthe control
the object
nature system
of have and
aa higher
the systems energy levelsupply
dynamics of requirements
complexity.
coupled withThe to
the Mirjalili (2015).
maintaining
Mirjalili (2015).
the The GWO
balance
The GWO
between algorithm
the
algorithm is capable
exploration
is capable
and of
the
of
levitate the object have higher level of complexity. The maintaining the balance between the
nonlinear
levitate thenature
object of the
have systems
a higher dynamics
level of coupled
complexity. with the
The exploitation
maintaining properties
the balance of the search
between the exploration
space. The
exploration and
andgreythe
the
(𝛼𝛼)
nonlinear
nonlinear nature
nature of
of the
characteristics
the systems
of thedynamics
actuatorscoupled complicate with the exploitation
maintaining properties
the balance of the
between search
the space.
exploration The and grey
the
the systems thedynamics coupled with the exploitation
wolves strictlyproperties
follow the of
of thesocial search
hierarchyspace.
space.as The
they greyare
(𝛼𝛼)
nonlinear characteristics
nature ofwas of the
systems actuators
dynamics complicate
coupled with the exploitation properties the search The grey
(𝛼𝛼)
nonlinear characteristics of actuators complicate the
(𝛽𝛽)
controller design consider by Marjan Golob et al. wolves strictly
exploitation follow
properties the
of social
the hierarchy
search space. as Thethey greyare

(𝛼𝛼)
nonlinear characteristics of the actuators complicate the wolves
categorisedstrictly
in to follow
four the
differentsocial hierarchy
groups namely, as they
alpha are
(𝛽𝛽)
controller design was consider by Marjan Golob et al.
(𝛼𝛼)
nonlinear characteristics of the actuators complicate the wolves strictly follow the social hierarchy as they are
as a(𝛽𝛽)
controller
(2003). For design
such a was
system consider
it is by
desired Marjan
to propose Golob a et
suitableal.
(𝛿𝛿) act (𝜔𝜔)
categorised
wolves in
strictly to four
follow different
the socialgroups namely,
hierarchy as alpha
they are

asZ.aa(𝛽𝛽)
controller
(2003). For design
such a was
system consider
it is by
desired Marjan
to propose Golob a et
suitableal. categorised
the leader, beta in to four
assistdifferent groups
alpha during namely,
the huntingalpha process,
(𝛿𝛿) act (𝛽𝛽) (𝜔𝜔)
controller
(2003).
controller For design
such a was
system consider
it is by
desired Marjan
to propose Golob a et
suitableal. categorised in to four different groups namely, alpha
Forfor positioning itaa isball in thetoair spaceawith the
(𝛿𝛿) the(𝜔𝜔)
(2003). For such a system
system it is desired propose suitable the leader,
categorised beta
in to assist
four alpha
different during
groups the hunting
namely, process,
alpha
(𝛿𝛿) (𝜔𝜔)
controller
(2003). for positioning
such a ball in
desired the to air space
propose awith the
suitable the
deltaleader, beta assist
subordinate alphaandduring
omega the hunting process,
the followers,
(𝛿𝛿) (𝜔𝜔)
controller
help of an for positioning
electromagnetic a ball
force. in
In the
the air
ideal space with
situation, the the
delta
the leader,
leader, actbeta
beta assist
assist alpha
subordinate alphaand during
omega
during the hunting
the
hunting process,
followers,
process,
controller
help of an
controller for
an for positioning aaforce.
electromagnetic
positioning ball in
ball in
Inthethe air
the current
the space
idealspace
air with the
situation,
with the delta
S. Saremi, S.as subordinate
Mirjalili, and andS. omega
M. Mirjalili the followers,
(2014). The
help of
magnetic electromagnetic
force producedforce. by In the ideal situation,
from the
an delta
S. Saremi,
delta act
actS.as
asZ.a
a subordinate
Mirjalili,
subordinate and and
andS. omega
M.
omega Mirjalili the
the followers,
(2014). The
followers,
help of
magnetic
help of an
an electromagnetic
force produced
electromagnetic force.
by
force. In
Inthethe
the ideal
current
ideal situation,
from
situation, the
an
the S. Saremi,
proposed S. Z.
algorithm Mirjalili,
tunes andthe S. M. Mirjalili
parameters (2014).
of the The
PID
magnetic
electromagnet force produced
counteracts the by the
weight current
of the from
metal an
ball. S. Saremi,
proposed
S. Saremi, S. Z.
algorithm
S. Z. Mirjalili,
tunes
Mirjalili, and
andthe S.
S. M. Mirjalili
parameters
M. Mirjalili (2014).
of the
(2014). The
PID
The
magnetic
electromagnet
magnetic force
force produced
counteracts
produced the by
by the
weight
the current
of the
current from
metal
from an
ball.
an proposed
controller algorithm
focuses on tunes
minimisingthe parameters
the integral of squarethe PID
error
electromagnet
However, counteracts
the electromagnetic the weight
force of
is very the metal
sensitive, ball.
and proposed
controller
proposed algorithm
focuses
algorithm on tunes
minimising
tunes the
the parameters
the integral
parameters of
square
of the
the PID
error
PID
electromagnet
However,
electromagnet the counteracts
electromagnetic
counteracts the
the weight
force
weight is of
very
of the
the metal
sensitive,
metal ball.
and
ball. controller
of the MLS. focuses
As the on minimising
error of thethe integral
system square
decreases, error
the
However,
there the
is noise electromagnetic
that creates acceleration force is very
forcessensitive,
on the ball, and controller
of the MLS.
controllerMLS. focuses
As the
focuses on
the minimising
error of
onspecifications
minimising thethe
of the the integral
system
integral square
decreases,
square error
the
error
However,
there
However, is thethe
noise
the electromagnetic
that creates
electromagnetic force
acceleration
force is
is very sensitive,
forcesregion,
very on the
theYadav
sensitive, and
ball,
and of the
desired time As
domain error system
are achieved decreases,
to meet the
there
causing is noiseballthat
to creates
move into acceleration
the unbalanced forces on ball, of the
the MLS.
desired
of MLS.
time As the
domain
As thespecifications
error of
error of the
theare system
achieved
system decreases,
to meet
decreases, the
the
there is noise that creates acceleration forces on the ball, desired time domain
causing
there
causing is the
noise
the
et al. (2012). ball to
that
ball to move
creates
move intointo the unbalanced
acceleration
the unbalanced forces region,
on theYadav
region, Yadav ball, system
desired
system
desired time domain specifications
requirements.
requirements.
time domain
The performance
specifications
The
specifications
are
are achieved
performance
are achieved
achieved
of to
of to
meet
tothe
meetPID
the
meet
the
the
PID
the
causing
et al. the
the ball
al. (2012).
causing (2012). ball to
to move
move intointo the the unbalanced
unbalanced region, region, Yadav
Yadav system
controller requirements.
is validated by The performance
analyzing of
the frequency the PID
plots.
et system
controller
system requirements.
isball
validated
requirements. The
by performance
analyzing
Theanalyzing
performance of
the frequency
frequency
of the
the PID
plots.
PID
et
et al.
In al. (2012).
this(2012).
paper, the metal ball is balanced in the air-space by controller
The metal is validated
of the by
MLS is the
levitated in the plots.
air-space
In this
this paper,
paper, the metal ball
ball is is balanced
balanced in the
the air-space by controller
The metal
controller is
is validated
ball of
validated the by
MLS
by analyzing
is
analyzing the
levitated
the frequency
in the
frequency plots.
air-space
plots.
In
controlling
In this paper,thethe
the metal
electromagnetic
metal ball is force of the
balanced in
in MLS
the air-space
using PID
air-space by
by The
within metal ball
fraction of ofthe MLS
seconds. is levitated
However, in the
the air-space
system
controlling
In this paper,the electromagnetic
the metal ball is force
balanced of the
in MLS
the using
air-space PIDby The
The metal
within metal ball
fraction
ball of
of ofofthe
the MLS
seconds.
MLS is
is levitated
However,
levitated in
in thethe air-space
the system
air-space
controlling
controller the
controlling the electromagnetic
because more thanforce
electromagnetic 90%
force of
of the
of MLS
the MLSthe using
using PID
industrial
PID within
performance fractionwhen theseconds.
parameters However,
are the
optimized system
using
controller
controlling because
the more
electromagnetic than 90%
force of of
the MLSthe industrial
using PID within
performance
within fraction
fractionwhen of
of theseconds.
parameters
seconds. However,
are
However, the
optimized
the system
using
system
controller
controllers
controller because
are still more
implemented than 90%
basedof the
around industrial
the PID performance
GWO algorithm when is the parameters
over-damped, are
which optimized
is acceptable using
for
controller because
controllers
controllers
algorithms are still
because
are
due still
to
more
implemented
more
implemented
its simplicity,
than
than 90% basedof
90%
based
ease
of the
the industrial
around
ofaround the PID
the
implementation PID
industrial performance
GWO
GWO
some algorithm
performance
algorithm
types of
when
when is
is
problems
the
the parameters
over-damped,
parameters
over-damped,
like, elevator,
are
which
are
which
optimized
is acceptable
optimized
is
liquid acceptable
level in
using
for
using
for
water
controllers
algorithms
controllers are still
due
are still
to implemented
its simplicity,
implemented basedof around
ease
based around the PID
implementation
the PID GWO
some
GWOand algorithm
types ofThe
algorithm is
problems
is over-damped,
like, and
over-damped, elevator,which
which is
is acceptable
liquid level
acceptablein showsfor
water
for
algorithms
and robustness.
algorithms due
due Theto
to its PID
its simplicity,
controllerease
simplicity, ease of
of implementation
is a combination
implementation of the some
tank types
etc. of problems
simulation like, elevator, liquid
experimental level
result in water
and robustness.
algorithms due The
to PID
its controller
simplicity, is
ease a combination
of implementation of the some
tank
some types
and etc.
types of
of problems
The simulation
problems like,
like, elevator,
and liquid
experimental
elevator, liquid level
result
level in
in water
shows
water
and
PI
and robustness.
and
robustness. The
PD controllers.
The PID
It iscontroller
PID similar tois
controller is aa combination
lag-lead
combination of
compensator.
of the
the tank
tank and
the etc.
etc. The
The simulation
effectiveness
and of the and
simulation experimental
proposed
and experimental result
result shows
controller, which
shows
PI
and
PI
Theand
andPI PDPD
robustness.controllers.
The
controllers.
control action It It
PID
and is
is similar
controller
similar
PD to
to
control islag-lead
a
lag-lead compensator.
combination
action compensator. of
occur in low the the
tank
the effectiveness
and etc. The
effectiveness
advances the system of performance
simulation
of the and
the proposed
experimental
proposed as controller,
result
controller,
compared withwhich
shows
whichthe
PI
The
PI and
andPI PD
PD controllers.
control action
controllers. It
and
It is
is similar
PD
similar to
control
to lag-lead
action
lag-lead compensator.
occur in
compensator. low the effectiveness
effectiveness
advances
the the system of performance
of the proposed
the proposed as controller,with
compared
controller, which
whichthe
The
and
The PI
high
PI control
frequencyaction and
regions. PD Thecontrol
PID controlaction occur
is usedin low
when advances
classical the
tuning system
approach. performance as compared with the
and
The
and
the PI control
high
high
system frequency
control
frequency
action
action
performance
and
regions. PD
andrequires
regions. PD The
The
control
PID
control
PID
action
control
action is
control
improvement
occur
is used
occur
used
in
in low
when
inwhen
both low
the
advancestuning
classical
advances
classical
the system
the
tuning
system
approach.
approach.
performance as
performance as compared
compared with with the the
and
the
and high
system
high frequency
performance
frequency regions.
requires
regions. The
The PID control
improvement
PID control is
is used
in both
used when
whenthe classical tuning approach.
classical tuning approach.
the
the system performance requires improvement in both the
the system
system performance
performance requiresrequires improvement
improvement in in both
both the the
2405-8963 ©
Copyright © 2016, IFAC (International Federation of Automatic Control)
2016 IFAC 778Hosting by Elsevier Ltd. All rights reserved.
Copyright
Peer review© 2016 responsibility
IFAC 778Control.
Copyright ©under
2016 IFAC of International Federation of Automatic
778
Copyright ©
© 2016
2016 IFAC
10.1016/j.ifacol.2016.03.151
Copyright IFAC 778
778
 IFAC ACODS 2016
February 1-5, 2016. NIT Tiruchirappalli, India
Shekhar Yadav et al. / IFAC-PapersOnLine 49-1 (2016) 778–782 779

𝑚𝑚𝐵𝐵 𝑥𝑥̃̈𝐵𝐵 = 𝑚𝑚𝐵𝐵 𝑔𝑔 − 𝑘𝑘 ̅𝑖𝑖̅𝑐𝑐 2 − 𝑘𝑘1 𝑖𝑖̃𝑐𝑐 + 𝑘𝑘2 𝑥𝑥̃𝐵𝐵

2
The organization of this paper is as follows. In Section 2,
𝑥𝑥𝐵𝐵
(7)

By defining the equilibrium point at 𝑥𝑥̅𝐵𝐵 = −1.5 volt,

Description of magnetic levitation system is presented. In

𝑖𝑖̅𝑐𝑐 = 0.8 A, the point where there is no acceleration and the

Section 3, the PID control technique is discussed. Section 4
presents the GWO algorithm. Section 5 covers results and
discussion. Section 6 presents the Conclusion and finally
mB g − k =0
𝑖𝑖̅𝑐𝑐 2
incremental terms are equal to zero i.e.,
the references.
𝑥𝑥̅𝐵𝐵 2
2. DESCRIPTION OF MAGNETIC LEVITATION

𝑥𝑥̃̈𝐵𝐵 = (𝑘𝑘2 𝑥𝑥̃𝐵𝐵 − 𝑘𝑘1 𝑖𝑖̃𝑐𝑐 )

The equation of motion takes the form of:
1
SYSTEM
𝑚𝑚𝐵𝐵
(8)
The MLS comprises of an electromagnet coil, a metal
ball and an IR position sensor. The metal ball sensed by an
=
Take the Laplace transform of equation (8), gives-
IR sensor separate the vertical movement of the ball from 𝑥𝑥̃𝐵𝐵 (𝑠𝑠) −𝐾𝐾1
𝑖𝑖̃ 𝑐𝑐 (𝑠𝑠) 𝑚𝑚𝐵𝐵 𝑠𝑠 2 −𝐾𝐾2
distance of the ball from the electromagnetic coil are 0.5
the horizontal one. The minimum and the maximum (9)

cm to 0.25 cm. The MLS can be decomposed into two The open-loop transfer function of the MLS is a second
subsystems, viz., a mechanical system and an electrical order system with complex conjugate poles for any value
system. The ball position in the mechanical system can be of the metal ball. The closed-loop response of the MLS is
controlled by adjusting the current through the oscillatory in nature i.e., either the metal ball attracted
electromagnet whereas the current in the electrical system towards the electromagnetic coil or it may be fallen down.
can be controlled by applying controlled voltage across the Therefore, there is a requirement of a controller which can
electromagnet terminals. Thus, the voltage applied across effectively achieve the desired performances.
the electromagnet terminals provides an indirect control of
the ball position has been discussed by C. R. Knospe and E. 3. PID CONTROLLER
G. Collins (1996). A continuous time PID controller is very popular since
The nonlinear model of the MLS is derived by using the 1900 era, as one of the earliest examples of a PID-type
fundamental principle of dynamics. The behaviour of the controller was developed by Elmer Sperry in 1911. Around
metal ball is given by the following electromechanical 90% of the industrial problems are handled by the PID
controllers. It is commonly considered as an extreme form

𝑚𝑚𝐵𝐵 𝑥𝑥̈ 𝐵𝐵 = 𝑚𝑚𝐵𝐵 𝑔𝑔 − 𝑓𝑓(𝑥𝑥𝐵𝐵 , 𝑖𝑖𝑐𝑐 )

equation:
of a phase lead-lag compensator with one pole at the origin
(1) and the other at infinity is discussed by Kiam Heong Ang
where f(xB , ic ) = kxic2 is the magnetic control force, mB is
2 and Gregory Chong (2005). The transfer function of the

the mass of the ball (20 × 10−3 kg) , g denotes the

B
PID controller is given by:

𝑈𝑈𝑃𝑃𝑃𝑃𝑃𝑃 (𝑠𝑠) = (𝐾𝐾𝑃𝑃 + 𝐾𝐾𝐼𝐼 + 𝐾𝐾𝐷𝐷 𝑠𝑠) 𝐸𝐸(𝑠𝑠)

1
gravitational constant (9.81 m/s 2 ), k is a constant related 𝑠𝑠
where 𝐾𝐾𝑃𝑃 is the proportional gain, 𝐾𝐾𝐼𝐼 is the integral gain,
(10)

coefficient, ic is the coil current and xB is the distance of 𝐾𝐾𝐷𝐷 is the derivative gain, 𝐸𝐸 is the error of the system and
to the mutual inductance of the ball and the coupling

the ball from the electromagnet. To levitate the ball in the 𝑈𝑈𝑃𝑃𝑃𝑃𝑃𝑃 is controller output. To find out the starting point of
air-space, the MLS should be linearized, as techniques like the PID controller, the parameters are tuned using ZN
root-locus, bode plot, pole-zero map and Nyquist plots are method due to its popularity and capability of providing the

linearized by taking the approximates of xB and ic as

not feasible for the nonlinear systems. The MLS is best result as compared with the other classical tuning
criterion.
𝑥𝑥𝐵𝐵 = 𝑥𝑥̅𝐵𝐵 + 𝑥𝑥̃𝐵𝐵 (2) 4. GREY WOLF OPTIMIZER
𝑖𝑖𝑐𝑐 = 𝑖𝑖̅𝑐𝑐 + 𝑖𝑖̃𝑐𝑐 (3) A Grey wolf (Canis lupus) optimizer is an evolutionary
where 𝑥𝑥̅𝐵𝐵 and 𝑖𝑖̅𝑐𝑐 are the value at the operating point, 𝑥𝑥̃𝐵𝐵 and
algorithm belongs to a Canidae family discussed by Muro
𝑖𝑖̃𝑐𝑐 shows the variations around the operating point. If we
assume that 𝑥𝑥̅𝐵𝐵 ≫ 𝑥𝑥̃𝐵𝐵 and 𝑖𝑖̅𝑐𝑐 ≫ 𝑖𝑖̃𝑐𝑐 and using Taylor’s series
C, Escobedo R, Spector L and Coppinger R (2011). The
grey wolf searches for the prey (food) in a group size of 5-
expansion gives: 12 wolves. The grey wolves strictly follow the social

𝑓𝑓(𝑥𝑥𝐵𝐵 , 𝑖𝑖𝑐𝑐 ) = 𝑘𝑘 ̅𝑖𝑖̅𝑐𝑐 2 + ( | 𝑖𝑖̅𝑐𝑐 + | 𝑥𝑥̅𝐵𝐵 ) (4)

2 𝜕𝜕𝜕𝜕(𝑥𝑥𝐵𝐵 ,𝑖𝑖𝑐𝑐 ) 𝜕𝜕𝜕𝜕(𝑥𝑥𝐵𝐵 ,𝑖𝑖𝑐𝑐 )
hierarchy and share the top level of hierarchy depending on
𝑥𝑥𝐵𝐵 𝜕𝜕𝑖𝑖𝑐𝑐 𝑖𝑖̅𝑐𝑐 ,𝑥𝑥̅𝐵𝐵 𝜕𝜕𝑥𝑥𝐵𝐵 𝑖𝑖̅𝑐𝑐 ,𝑥𝑥̅𝐵𝐵
its dominating behaviour. The top level represents the

to this level is known as the leader denoted as alpha (α).

dominating level of the hierarchy, and the one who belongs
From the above equation, the higher order terms are
neglected and the first order approximation is chosen. The

wolf who assist α during the hunting process commonly

partial differential terms can be solved as: The second level of the hierarchy is being shared by the

| = −2𝑔𝑔/𝑖𝑖̅𝑐𝑐 = −𝑘𝑘1 known as beta (β). The third level of hierarchy is being
𝜕𝜕𝜕𝜕(𝑥𝑥𝐵𝐵 ,𝑖𝑖𝑐𝑐 )
𝜕𝜕𝑖𝑖𝑐𝑐
shared by delta (δ) wolf, which act as a subordinate. The
𝑖𝑖̅𝑐𝑐 ,𝑥𝑥̅𝐵𝐵
(5)

| = 2𝑔𝑔/𝑥𝑥̅𝐵𝐵 = 𝑘𝑘2
𝜕𝜕𝜕𝜕(𝑥𝑥𝐵𝐵 ,𝑖𝑖𝑐𝑐 )
𝜕𝜕𝑥𝑥𝐵𝐵 𝑖𝑖̅𝑐𝑐 ,𝑥𝑥̅𝐵𝐵
(6) last level is the least dominating level of the hierarchy and

known as omega (ω) . The exploration property is

the wolves who represent this level are the followers
Substituting equation (5) and (6) in to equation (1)

779
IFAC ACODS 2016
780 Shekhar Yadav et al. / IFAC-PapersOnLine 49-1 (2016) 778–782
February 1-5, 2016. NIT Tiruchirappalli, India

preserved by splitting the searching criterion into three undamped and therefore, to control such an oscillatory
different directions is included by Emary, E., et al. (2015). system, the controller should be fast and capable of

1. Maximum Overshoot ≤ 15%

The GWO algorithm comprises of following steps: improving the design requirements such as:

A. Encircling the Prey 2. Settling Time ≤ 1 𝑠𝑠𝑠𝑠𝑠𝑠.

The grey wolves encircle the prey by updating their The parameters of the PID controller are obtained using the
PID tool of MATLAB and tuned by Z-N tuning rule. The
𝐷𝐷 = |𝐶𝐶. 𝑋𝑋𝑃𝑃 (𝑡𝑡) − 𝑋𝑋(𝑡𝑡)|
position with respect to the top level of the hierarchy.
PID controller improves the settling time of the system but
(11)
𝑋𝑋(𝑡𝑡 + 1) = 𝑋𝑋𝑃𝑃 (𝑡𝑡) − 𝐴𝐴. 𝐷𝐷
unable to obtain the desired overshoot as shown in Table 1.
(12) The step response of the MLS with PID controller is shown
𝐴𝐴 = 2𝑎𝑎. 𝑟𝑟1 − 𝑎𝑎
in Fig. 1.
(13)
𝐶𝐶 = 2. 𝑟𝑟2
Table 1. Performance Characteristics with PID controller

where A and C are coefficient vectors, XP is the position

(14)
Gain 𝑲𝑲𝑷𝑷 Gain 𝑲𝑲𝑰𝑰 Gain 𝑲𝑲𝑫𝑫
Proportional Integral Derivative Settling Max.

vector of the prey, X indicates the position vector of the (𝒕𝒕𝒔𝒔 )

Time Overshoot

grey wolf, t is the current iteration, r1 and r2 are the 9.4085 19.8483 0.05730 0.5030 37.6922
random values lie in the range [0, 1].
B. Hunting Behaviour

considered as α, β, and δ. The positions of the other wolves

The wolves which provides the best solutions were

are reorganized with the correspondence to these three

wolves as:
𝐷𝐷𝛼𝛼𝑖𝑖 = |𝐶𝐶1 . 𝑋𝑋𝛼𝛼𝑖𝑖 (𝑡𝑡) − 𝑋𝑋𝑖𝑖 (𝑡𝑡)|
𝐷𝐷𝛽𝛽𝑖𝑖 = |𝐶𝐶2 . 𝑋𝑋𝛽𝛽𝑖𝑖 (𝑡𝑡) − 𝑋𝑋𝑖𝑖 (𝑡𝑡)| }
𝐷𝐷𝛿𝛿𝑖𝑖 = |𝐶𝐶3 . 𝑋𝑋𝛿𝛿𝑖𝑖 (𝑡𝑡) − 𝑋𝑋𝑖𝑖 (𝑡𝑡)|
(15)

𝑋𝑋𝑖𝑖1 = 𝑋𝑋𝛼𝛼𝑖𝑖 (𝑡𝑡) − 𝐴𝐴1 . 𝐷𝐷𝛼𝛼𝑖𝑖

𝑋𝑋𝑖𝑖2 = 𝑋𝑋𝛽𝛽𝑖𝑖 (𝑡𝑡) − 𝐴𝐴2 . 𝐷𝐷𝛽𝛽𝑖𝑖 }
𝑋𝑋𝑖𝑖3 = 𝑋𝑋𝛿𝛿𝑖𝑖 (𝑡𝑡) − 𝐴𝐴3 . 𝐷𝐷𝛿𝛿𝑖𝑖
(16)

𝑋𝑋𝑖𝑖 (𝑡𝑡 + 1) =
𝑋𝑋𝑖𝑖1 +𝑋𝑋𝑖𝑖2 +𝑋𝑋𝑖𝑖3
3
(17)
where i is the no. of iterations and Xi (t + 1) is the best
search agent for the iteration i. Fig. 1. Step response of MLS with PID controller

C. Attacking Prey The comparison of closed-loop step response of MLS with

wolves and the location of the prey, the coefficient vector A

To reduce the gap between the position of the grey and without PID controller is not possible because the
original system is oscillatory. To show the effectiveness of

minimizes, the value of vector A is decreased as

is decreased. As the mean of the performance index the PID controller, the frequency plot are analysed as
shown in Fig. 2.

𝑎𝑎 = 2 − ( )
2
𝑚𝑚𝑚𝑚𝑚𝑚.𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖
(18)

where 𝑚𝑚𝑚𝑚𝑚𝑚. 𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 is the maximum value of the performance

index at iteration 𝑖𝑖, 𝑎𝑎 is decreased from 2 to 0 as the no.
of iteration increases.
5. RESULTS & DISCUSSION
In this paper, a PID controller is designed for MLS to
levitate the metal ball in the air-space. The MLS consists of
electromagnetic coil, mounted on top of the box. The IR
sensor is fixed on opposite sides of the box, so that it may
easily sense the position of the metal ball. The metal ball
gets attracted when it is near to the electromagnetic coil
and otherwise fallen down due to the gravitational force.
The purpose of the PID controller is to maintain an
equivalent electromagnetic force, so that the metal ball is

𝑠𝑠 = ±𝑗𝑗43.3590 . The step response of the MLS is

levitated. The closed-loop poles of the MLS are locates at
Fig. 2. Bode plot of MLS with and without controller

780
 IFAC ACODS 2016
February 1-5, 2016. NIT Tiruchirappalli, India
Shekhar Yadav et al. / IFAC-PapersOnLine 49-1 (2016) 778–782 781

To meet all the design requirements the parameters of the The bode plot shown in Fig. 5 to compare both the tuning
PID controller are optimized using GWO algorithm. The criterion. The bode response validate the effectiveness of
GWO algorithm optimizes the parameters of the PID the GWO algorithm. The parameters of PID controller
controller while minimizing the performance indices like optimized using GWO algorithm is shown in Table 2.

𝐽𝐽𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 = ∫0 𝑡𝑡. |𝑒𝑒(𝑡𝑡)| 𝑑𝑑𝑑𝑑

∞
ITAE and ITSE.
(19)

𝐽𝐽𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 = ∫0 𝑡𝑡. 𝑒𝑒 2 (𝑡𝑡) 𝑑𝑑𝑑𝑑

∞
(20)
The search space for the GWO algorithm is selected on the
basis of the parameters tuned using Z-N rule and it will act
as a leader for the first iteration. The other wolves are
selected randomly around alpha by a suitable guess. The
search space should be defined properly to obtain a global

controller is defined as follows: (1 ≤ 𝐾𝐾𝑃𝑃 ≤ 350) ,

optimum solution. The range for the parameters of PID

(0.01 ≤ 𝐾𝐾𝐼𝐼 ≤ 300) and (0.0001 ≤ 𝐾𝐾𝐷𝐷 ≤ 200) . The

optimized parameters of PID controller are obtained by
searching for the global optimum solution for around 500
of iterations. The comparison of step response of the MLS
with PID controller tuned using GWO and Z-N tuning rule
is shown in Fig. 3.
Fig. 5. Bode plot of MLS with PID tuned using GWO & Z-N rule

Table 2. Parameters optimized using GWO algorithm

Grey wolves 𝐊𝐊 𝐏𝐏 𝐊𝐊 𝐈𝐈 𝐊𝐊 𝐃𝐃
α 222.8579 181.8144 33.9586

𝛅𝛅
β 290.0806 236.6568 44.2018

𝛚𝛚𝟏𝟏
173.7181 141.7246 26.4708

𝛚𝛚𝟐𝟐
242.4454 197.7945 36.9433

𝛚𝛚𝟑𝟑
278.2969 227.0433 42.4063

𝝎𝝎𝟒𝟒
158.9236 129.6548 24.2164

𝝎𝝎𝟓𝟓
278.2673 227.0191 42.4017

𝝎𝝎𝟔𝟔
260.6799 212.6708 39.7218

𝝎𝝎𝟕𝟕
316.2232 257.9847 48.1854

𝝎𝝎𝟖𝟖
54.9622 44.8399 8.3750
Fig. 3. Step response of MLS with PID tuned using GWO & Z-N rule

𝝎𝝎𝟗𝟗
224.2803 182.9749 34.1753
As the settling time of the system is very fast therefore, 190.9889 155.8147 29.1025
responses of alpha, beta and delta are shown in Fig. 4.

The performance characteristics of MLS with GWO

algorithm in terms of time domain specification is shown in
Table 3. The performance indices like ITAE and ITSE are
reduced to certain extent.
Table 3. Comparison of performance characteristics
Settling Max. ITAE ITSE
time Overshoot
(seconds) (%)
MLS- 0.000140 0.0202 0.5405 0.0006
ALPHA
MLS-BETA 0.000143 0.0207 0.5514 0.0006

MLS- 0.000144 0.0208 0.5556 0.0006

DELTA
Fig. 4. Step response of MLS with PID tuned using GWO MLS-ZN 0.5030 37.6922 1.3769 0.0269

781
IFAC ACODS 2016
782 Shekhar Yadav et al. / IFAC-PapersOnLine 49-1 (2016) 778–782
February 1-5, 2016. NIT Tiruchirappalli, India

6. CONCLUSION

An optimized PID controller is designed to levitate the

metal ball of the MLS. The simplified mathematical
model has been developed and the transfer function of
the MLS has been established. The parameters of the
conventional PID controller are optimized by using a meta-
heuristic algorithm known as GWO. The proposed
optimization technique ensures the improvement of the
time domain and as well as the frequency domain
specifications by minimising the performance indices of
the system. The performance index chosen in this paper are
ITAE and ITSE. The eminent properties of optimization
technique like exploration and exploitation are guarantees
by the GWO algorithm. The effectiveness of the proposed
techniques is validated by comparing the results with the
classical design techniques.

REFERENCES
[1] C. R. Knospe and E. G. Collins (1996). “Special Issue on
Magnetic Bearing Control,” IEEE Trans. Control System
Technology, Vol. 4, No. 5.
[2] Rule R. and Gilliland R. (1980). “Combined magnetic
levitation and propulsion: The mag-transit concept,” IEEE
Trans. Veh. Technol. VT-29, 41–49.
[3] Ion B., Trica A., Papusoiu G., and Nasar A. S. (1988).
“Field tests on a maglev with passive guideway linear
inductor motor transportation system,” IEEE Trans. Veh.
Technol. VT-38, 230–236.
[4] Vinodh Kumar E, Jovitha Jerome (2013). “LQR based
optimal tuning of PID controller for trajectory tracking of
Magnetic Levitation System,” Procedia Engineering 64,
254 – 264.
[5] Marjan Golob, Boris Tovornik (2003). “Modeling and
control of the magnetic suspension system,” ISA
Transaction 42, 89-100.
[6] Shekhar Yadav, J. P. Tiwari, S. K. Nagar (2012). “Digital
Control of Magnetic Levitation System using Fuzzy Logic
Controller,” International Journal of Computer
Applications (0975 – 8887), Vol. 41– No.21.
[7] Katsuhiko Ogata (2009). “Modern Control Engineering,”
4th ed., Pearson Education Inc., Upper Saddle River, New
Jersey 07458, U.S.A.
[8] S. Mirjalili, S. M. Mirjalili, and A. Lewis. (2014). “Grey
Wolf Optimizer,” Advances in Engineering Software, Vol.
69, pp. 46-61.
[9] S. Saremi, S. Z. Mirjalili, and S. M. Mirjalili (2014).
“Evolutionary population dynamics and grey wolf
optimizer,” Neural Computing and Applications: 1-7.
[10] Seyedali Mirjalili (2015). “How effective is the Grey Wolf
optimizer in training multi-layer perceptrons,” Applied Intel.
1-12.
[11] Muro C, Escobedo R, Spector L and Coppinger R (2011).
“Wolf-pack (Canis lupus) hunting strategies emerge from
simple rules in computational simulations,” Behav Process
88(3):192-7.
[12] Emary, E., et al. (2015). “Feature Subset Selection
Approach by Gray-Wolf Optimization,” Afro-European
Conference for Industrial Advancement. Springer
International Publishing.
[13] Kiam Heong Ang and Gregory Chong (2005). “PID control
system analysis, design and technology,” IEEE Transaction
on Control System Technology, Vol. 13, No. 4, July.

View publication stats

782