Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
13
,
No.
2
,
F
eb
r
uar
y
201
9
, pp.
4
75
~484
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v1
3
.i
2
.pp
4
75
-
484
475
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Wind
directi
on senso
r
b
ased
on th
er
m
al anemom
ete
r for
olfacto
ry mobile
robot
Helm
y Wi
dya
nt
ar
a
,
Muha
mmad
R
i
va
i
,
Djok
o
P
urw
anto
Depa
rtment
o
f
E
le
c
tri
c
al E
ngin
eering,
Inst
it
ut
T
ek
nologi
Sepu
luh N
opember
Suraba
y
a, I
ndonesi
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Sep
1
9
, 201
8
Re
vised
N
ov
2
0,
2018
Accepte
d
Dec
4
, 2
018
A
wind
dire
ct
i
on
sensor
has
bee
n
implement
ed
for
m
an
y
a
ppli
c
at
ions
,
such
as
navi
g
at
i
on,
wea
the
r
,
and
ai
r
po
ll
uti
on
m
onit
oring
.
In
an
o
dor
tra
ck
ing
s
y
stem,
th
e
win
d
play
s
the
important
ro
le
to
ca
rr
y
g
as
from
it
s
source
.
The
ref
or
e,
th
e
pre
ci
se
,
low
-
cos
t,
and
eff
e
ctive
wind
dire
ct
ion
sensor
is
req
uire
d
to
tr
ac
e
the
gas
sourc
e
.
In
thi
s
stud
y
,
a
n
ew
design
of
wi
nd
dire
c
ti
on
sensor
has
bee
n
deve
lop
ed
usin
g
the
rm
al
ane
m
om
et
er
prin
ci
pl
e
with
the
m
ai
n
component
of
the
p
ositi
ve
te
m
per
at
u
re
coe
f
ficien
t
the
rm
istor.
Thre
e
an
emom
et
ers
each
of
whic
h
has
diff
ere
n
t
d
ire
c
ti
ons
ar
e
use
d
as
i
npu
ts
for
the
neur
al
ne
twork
to
det
ermine
the
direct
ion
of
the
wind
aut
om
at
ic
a
l
l
y
.
The
expe
r
iment
al
r
esult
s
show
t
hat
the
w
ind
sen
sor
s
y
stem
is
ab
le
to
d
et
e
ct
twel
ve
wind
dir
e
ct
ions.
A
m
obil
e
robot
equi
pped
with
thi
s
sensor
s
y
stem
can
navi
ga
te
to
a
wind
so
urc
e
in
the
open
ai
r
with
a
succ
ess
rat
e
of
80%
.
Thi
s
s
y
stem
is
expe
cted
to
inc
r
e
ase
the
succ
ess
rat
e
of
th
e
olfac
tor
y
m
obile
robot
in
sea
r
chi
n
g
for
d
ange
rous l
ea
king
gas
in
th
e
open ai
r
.
Ke
yw
or
ds:
Neural
netw
ork
Olfacto
ry m
ob
il
e rob
ot
Ther
m
al
an
em
om
et
er
W
i
nd
directi
on
sen
s
or
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed.
Corres
pond
in
g
Aut
h
or
:
Muh
am
m
ad
Riv
ai
,
Dep
a
rtm
ent o
f El
ect
rical
En
gi
neer
i
ng,
In
sti
tut Te
knol
og
i
Sepulu
h N
op
em
ber
Surab
ay
a, Indo
nesia
.
Em
a
il
:
m
uh
a
m
m
ad_
ri
vai@ee.
it
s.ac.id
1.
INTROD
U
CTION
The
m
easur
e
m
ents
of
wind
directi
on
a
re
com
m
on
ly
us
e
d
in
we
at
her
m
on
it
ori
ng
sta
ti
ons
,
nav
i
gation
syst
e
m
s,
an
d
ai
r
poll
ution
syst
e
m
s
[1,
2]
.
T
his
data
is
re
qu
i
red
in
m
any
ar
eas,
s
uch
as
a
viati
on,
sh
ip
ping,
ag
ric
ultur
e
,
et
c.
I
n
po
ll
utio
n
syst
em
s,
the
wind
directi
on
is
use
d
for
analy
sis
of
the
sp
rea
d
of
ai
r
po
ll
uta
nts
[3,
4]
,
especial
ly
in
industrial
area
with
neig
hbori
ng
reside
nc
e.
W
in
d
di
recti
on
is
al
so
use
d
as
aux
il
ia
ry
de
vic
e
to
m
ake
the
ta
keo
f
f
pro
ce
ss
of
Un
m
anned
Ae
rial
Vehi
cl
e
(U
A
V)
s
m
oo
ther
[
5]
,
and
as
a
nav
i
gation sy
stem
o
n o
dor
tra
ckin
g rob
ot
[
6]
.
In
the
od
or
tra
ckin
g
rob
ot,
th
e
wind
play
s
a
ro
le
to
carry
the
od
or
fro
m
i
ts
so
ur
ce
[
3].
The
odor
m
ov
es f
ro
m
th
e so
urce occur
s b
y adv
ect
io
n
and
diffusio
n
a
ssist
ed
by the w
in
d.
Th
e
odor p
lum
e sp
reads t
o
fill
the
spa
ce
f
ollo
wing
the
wi
nd
directi
on
as
il
lustrate
d
in
Fig
ur
e
1
[
7].
Se
ve
ral
m
et
ho
ds
of
odor
pl
um
e
trackin
g
hav
e
bee
n
te
ste
d
an
d
com
pared
it
s
perform
ance
by
pre
vious
resea
rc
her
s
.
Li
et
al
con
cl
ud
e
d
that
the
z
igzag
m
et
ho
d
was
ef
fici
ent
only
if
the
r
obot
m
ov
es
faster
tha
n
the
odor
pl
ume
or
ai
r
flo
w.
The
upwind
m
et
ho
d
cou
l
d
be
bette
r
if
the
ro
bot
m
ov
e
s
slow
e
r
th
an
the
odor
pl
um
e
[
3
]
.
Ho
w
ever,
the
pr
e
vi
ou
s
pro
po
se
d
m
et
ho
d
has
no
t been
s
uccess
fu
l
in
tr
a
ci
ng
t
he
odor pl
um
e
in
open
a
ir
an
d
in
tu
rbul
ent
wind
c
ondi
ti
on
s
. Unstable
wind
conditi
ons
cau
se
the
dire
ct
io
n
of
s
pr
ea
ding
the
odor
plu
m
e
is
change
d,
t
his
cause
s
the
rob
ot
to
lose
tr
ackin
g
of
t
he
od
or
plum
e.
Ther
e
fore,
the
searc
h
for
the
od
or
s
our
ce
is
unsu
cces
sfu
l
or
ta
kes
a
long
ti
m
e
to
find
t
he
odor
pl
um
e.
Alexa
nd
e
r
et
al
con
si
der
e
d
the
wi
nd
dire
ct
ion
in
his
m
et
ho
d
for
tr
ackin
g
odor
pl
um
e
in
tur
bu
le
nce
c
onditi
on
[
8
]
.
T
he
trackin
g
s
ucce
ss
rate
i
ncr
ease
s
to 8
0%
in
the
sim
ulatio
n.
T
hi
s
resu
lt
em
ph
a
siz
es
that
the
pr
es
en
t
of
wi
nd
direc
ti
on
se
nsor
is
necessa
ry
f
or
odor
plu
m
e
trackin
g
in
real
pl
an
ts.
By
knowing
th
e
wind
di
recti
on,
the
trac
king
process
ca
n
be
accom
plished
faster
a
nd
eas
ie
r
to
fin
d
t
he
gas
s
ource
.
I
n
this
stud
y,
we
pro
pose
a
wind
dir
ect
ion
se
ns
or
usi
ng
an
om
nid
irect
ion
al
t
her
m
al
anem
o
m
et
er
.
T
his
sen
sor
s
yst
e
m
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
2
,
Fe
bru
ary 2
019
:
475
–
484
476
is
exp
ect
e
d
to
be
im
ple
m
ented
in
var
io
us
un
m
ann
e
d
de
vi
ces,
su
c
h
as
odor
tracki
ng
rob
ot
or
gas
l
eakag
e
fin
der
r
obot
[
9,
10]
, fo
r
ei
ther
sing
le
[11]
or
s
war
m
ty
pe
rob
ots
[12
-
14]
.
Figure
1.
The
odor
plu
m
e d
ispersi
on
2.
RESEA
R
CH MET
HO
D
2
.
1.
Ther
m
al
A
nem
ome
ter
An
em
om
e
te
r
is
a
de
vice
to
m
easur
e
the
w
ind
s
pee
d.
Ba
s
ed
on
it
s
w
ork
ing
pr
i
nciple,
there
are
t
w
o
ty
pes
of
this
dev
ic
e,
i.e.
ve
locit
y
and
pr
e
ssu
re
anem
ome
te
rs.
H
owev
e
r,
the
vel
ocity
anem
o
m
et
er
i
s
m
or
e
com
m
on
ly
us
ed
in
m
any
appl
ic
at
ion
s.
T
herm
al
ane
m
o
m
eter
is
a
ty
pe
of
velocit
y
anem
om
et
er
that
con
ve
rts
wind
s
peed
to
tem
per
at
ur
e
ch
ang
e
s.
In
this
m
et
ho
d,
w
he
n
the
ai
r
is
passi
ng
thr
ough
t
he
wire
or
tra
nsd
ucer’s
su
r
face,
it
will
in
du
ce
the
co
oling
ef
fect
[15]
.
T
he
dif
fer
e
nt
m
agn
it
ud
e
of
the
te
m
per
at
ur
e
is
the
n
c
onve
rte
d
into
wind
spe
e
d.
T
her
m
al
ane
m
o
m
et
er
has
hig
h
f
reque
ncy
respo
ns
e
an
d
good
sp
at
ia
l
res
olu
ti
on
com
pared
to
the o
t
her m
et
h
od
s
. T
he
refor
e
,
it
is b
r
oa
dly u
sed for t
urbu
le
nt w
i
nd
analy
si
s
.
The
therm
al
anem
o
m
et
er
i
s
sh
own
in
F
igure
2.
Ge
ne
rall
y,
the
core
el
e
m
ent
of
the
therm
a
l
anem
o
m
et
er
is
a
thin
wire
know
n
as
ho
t
wire
or
a
t
ransduce
r
s
uc
h
as
a
therm
ist
or
,
resist
ive
te
m
p
erature
detect
or
s
(RT
D)
,
et
c.
The
therm
al
anem
o
m
et
er
has
ge
ne
r
al
w
orki
ng
pri
nciple
i.e
.,
w
hen
the
tran
sducer
i
s
su
ppli
ed
by
a
n
el
ect
rical
cu
rr
e
nt,
it
will
gen
e
rate
the
i
nter
nal
heati
ng
w
hich
e
qual
s
to
it
s
su
r
rou
nd
i
ngs
[16,
17]
.
T
he
a
m
ou
nt
of
i
nput
powe
r
is
e
qu
al
to
t
he
lo
st
power
a
ff
ec
te
d
by
the
c
onvecti
ve
heat
tran
s
fer
expresse
d
as:
2
=
ℎ
.
(
−
)
(1)
W
he
re
I
is
the
inp
ut
cu
rr
e
nt,
R
w
is
the
resist
ance
of
the
t
ran
s
ducer
,
h
is
the
heat
trans
fer
coe
ff
ic
ie
nt
of
the
trans
du
ce
r
,
A
w
is
the
trans
duc
er’
s
s
urface
a
r
ea,
T
w
is
the
trans
du
ce
r
te
m
p
eratur
e
,
T
f
is
the
te
m
per
at
ur
e
in
the
ai
r,
a
nd
R
w
is t
he fu
nctio
n of
t
e
m
per
at
ure
deri
ved
a
s:
=
[
1
+
α
(
−
)
]
(2)
W
he
re
is
the
coeffic
ie
nt
of
therm
al
resist
a
nce,
T
Ref
is
the
ref
e
ren
ce
te
m
per
at
ur
e
,
a
nd
R
R
ef
is
the
resist
ance
at
the
ref
e
ren
ce
t
e
m
per
at
ure
[18
]
.
A
ccordin
g
to
King'
s
la
w
,
the
heat
trans
fe
r
coeffic
ie
nt
sym
bo
li
zed
by
h
is
the
functi
on
of
flui
d velocit
y
V
f
,
wh
e
rein
a
,
b
, a
nd
c
a
re th
e
coe
ff
ic
ie
nt
ac
hie
ved
f
ro
m
the c
al
ibrati
on (
c
~
0.5).
ℎ
=
+
.
(3)
By
co
m
bin
ing
the equat
io
n (1), (2
), an
d (3),
we
ca
n
el
im
ina
te
h
.
+
.
=
2
(
−
)
=
2
[
1
+
(
−
)
]
(
−
)
(4)
Ther
e
f
or
e,
the
fluid vel
ocity
can be
wr
it
te
n
a
s:
=
{
[
2
[
1
+
(
−
)
]
(
−
)
−
]
/
}
1
/
(5)
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Wi
nd d
ire
ct
io
n sens
or
base
d on ther
m
al ane
mometer f
or
olfactory
m
ob
il
e
ro
bot
(
Hel
my W
idya
nta
r
a
)
477
Figure
2
.
The
r
m
al
an
em
o
m
e
t
er: (a
)
the
workin
g pr
i
nciple,
(b) W
i
nd Re
vP sens
or, a
nd
(c)
c
ha
racteri
sti
c o
f
the
Win
d R
evP
se
nsor
2
.
2
.
Win
d Tu
nnel
The
wind
t
unne
l
is
us
e
d
t
o
m
ini
m
iz
e
the
eff
ect
s
of
tu
rbulence,
to
m
ain
ta
in
th
e
sta
bl
e
ai
rf
lo
w,
a
nd
al
so
to
pr
oduc
e
the
hom
og
en
eous
ai
rf
l
ow
[
19,
20]
.
T
he
wind
tu
nn
el
is
app
li
ed
t
o
ac
hieve
the
wind
sens
or
char
act
e
risti
cs,
an
d
al
s
o
to
ve
rify
the
ex
per
i
m
ental
resu
lt
s.
The
desi
gn
of
the
wind
t
unne
l
is
show
n
i
n
Figur
e
3
.
T
he
wind
t
unne
l
us
e
d
in
th
is
exp
e
rim
ent
consi
sts
of
a
c
ham
ber
with
t
he
siz
es
of
15c
m
x
18
cm
x
18cm
.
It
has
the
e
ntry
a
nd
t
he
e
xit
ho
l
e
s
with
the
sa
m
e
dia
m
et
er
of
18
cm
.
Ther
e
are
six
t
hin
fins
instal
le
d
at
the
exit
ho
le
to
m
ini
m
i
ze
the
tur
bule
nc
e
eff
ect
.
F
urt
her
m
or
e,
th
e
di
ff
use
r
of
plasti
c
straw
locat
e
d
at
the
ent
ry
channe
l
has
the
le
ngth
of
15
cm
and
diam
e
te
r
of
0.
7
cm
.
The
ai
r
com
ing
from
t
he
ou
tsi
de
al
re
ady
recti
fied
by
the
diffuse
r
is
tran
sp
ort
ed
t
o
the
te
sti
ng
area
with
the
siz
e
of
0.6m
x
0.
4m
x
0.
4m
.
In
this
stu
dy
,
the
wi
nd
t
unne
l
op
e
rates
from
0.5 to 7
m
ph
, wit
h
the
volt
ag
e
sup
ply
is va
ried fr
om
7
.5
vo
lt
s to
24
volt
s.
Figure
3
.
The
desig
n of wi
nd tunnel
2
.
3
.
Win
d
Direction
Sens
or
This
stu
dy
us
e
s
a
therm
al
anem
o
m
et
er
based
wind
s
pee
d
se
ns
or
of
W
i
nd
Re
vP
pro
duce
d
by
Mo
dern
Dev
ic
e.
It
ap
pl
ie
s
a
posit
ive
te
m
per
at
ur
e
coeffic
ie
nt
(PTC
)
the
rm
ist
or
as
the
tran
s
du
ce
r
t
o
m
easur
e
the
coo
li
ng
e
ff
ect
cause
d
by
the
ai
r
passing
th
r
ough
the
tra
nsdu
ce
r’s
surface
.
It
is
able
to
m
easur
e
t
he
wi
nd
sp
ee
d
from
0
to 150
m
ph
[
2
1],
with
the acc
ur
acy
of 0.5
m
ph
[2
2]. F
urt
he
rm
or
e, the
relat
io
nship
b
et
wee
n
t
he v
oltage
and tem
per
at
ure can
be
e
xpres
sed
as:
=
+
+
(6)
wh
e
re
T
is
a
m
bient
tem
per
at
ur
e
(
C),
is
wind
sp
ee
d
(m
s
-
1
),
an
d
a,
b,
c,
d
are
co
ns
ta
nts.
T
he
tra
nsp
os
e
of
wind s
pee
d
ca
n be
wr
it
te
n
as:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
2
,
Fe
bru
ary 2
019
:
475
–
484
478
=
{
(
−
−
)
}
1
/
(7)
The
outp
ut
sig
nal
of
t
he
Wi
nd
Re
vP
is
t
he
n
c
onve
rted
into
dig
it
al
data
us
i
ng
A
nalo
g
to
Digital
Converte
r
(
A
DC)
of
the
m
i
cro
c
ontrolle
r
.
The
c
har
act
eri
sti
c
data
of
th
e
W
in
d
Re
vP
assessed
by
th
e
wind
tunnel
can
be
seen
in
Fi
gure
2(
c
).
It
s
hows
that
the
wind
sens
or
pro
vide
s
the
dif
fer
e
nt
respo
ns
e
s
f
or
each
sens
or
to
the
wind
directi
on
.
This
data
is
per
f
orm
ed
us
ing
a
wi
nd
s
peed
of
1.5,
3,
an
d
6
m
ph
with
norm
al
iz
a
ti
on
expresse
d
as:
=
(
−
)
(
−
)
(8)
Accor
ding
to
e
xp
e
rim
ental
resu
lt
,
the
wind
sens
or
has
the
hig
hest
ou
t
pu
t
at
the
di
recti
on
of
0
a
nd
the
lo
west
out
pu
t
at
t
he
di
re
ct
ion
of
90
,
a
nd
27
0
.
Me
a
nwhile
,
f
or
the
opposit
e
di
re
ct
ion
s
uch
as
0
a
nd
180
,
the
wind
sen
sor
pr
ov
i
des
sim
i
la
r
outpu
t.
T
her
e
f
ore,
to
obta
in
th
e
pr
eci
se
wind
directi
on
m
od
el
ing,
an
a
d
va
nce
d
m
et
hod
a
nd app
r
opriat
e sen
sor
placem
ent are
require
d.
Accor
ding
to
t
he
sen
sor
cha
r
act
erist
ic
,
the
wind
sen
sor
ha
s
a
good
se
nsi
ti
vity
in
the
m
easur
em
ent
ang
le
of
60
.
I
n
orde
r
to
rec
ognize
12
wi
nd
directi
ons,
the
config
ur
at
io
n
of
th
ree
sens
ors
def
ine
d
with
A,
B,
and
C
s
houl
d
be
locat
e
d
at
0
,
12
0
,
a
nd
240
,
resp
ect
i
ve
ly
.
The
se
ns
or
config
ur
at
io
n
i
s
show
n
in
Fig
ur
e
4.
The
sm
al
l
siz
e
of
a
nem
o
m
et
ers
are
r
eq
uire
d
f
or
m
ob
il
e
r
obot
app
li
cat
ions
s
uch
as
s
war
m
ro
boti
cs.
Accor
ding
to
the
m
echani
cal
design
,
if
the
base
diam
et
er
of
the
wi
nd
directi
on
sens
or
is
sm
a
ller
than
20
cm
,
it
wil
l
gen
e
rate t
he
si
m
il
ar r
esp
on
se
s due t
o
the
shor
t
distance.
The
sc
hem
at
ic
la
yout
of
the
wind
di
recti
on
sens
or
s
how
n
in
Fig
ur
e
4(
a
)
us
es
the
om
nid
irect
io
nal
m
et
ho
d
a
pply
ing
t
he
ac
r
yl
ic
m
at
erial
with
a
thick
ness
of
0.5
cm
.
Figure
4(b)
prese
nts
the
m
echan
ic
al
desig
n
of
t
he
wi
nd
di
recti
on
se
nsor
.
Each
se
nsor
i
s
co
nn
ect
e
d
to
the
m
ic
ro
con
t
ro
ll
er.
The
dat
a
retrieval
pro
cess
is
cond
ucted
us
i
ng
the
wi
nd
tunnel
at
t
he
w
ind
sp
ee
d
of
1.5
m
ph,
3
m
ph
,
a
nd
6
m
ph
for
t
he
s
pecifi
c
wi
nd
directi
ons
by rotat
ing
t
he
se
nsor
ev
e
ry
30
.
Figure
4.
W
i
nd d
irect
io
n
se
nsor:
(a) t
he
sc
he
m
at
ic
lay
ou
t, and (b)
the m
echan
ic
al
desig
n
2
.
4
.
Neur
al
Net
w
or
k
fo
r
Wind
Directio
n Predicti
on
In
this
stu
dy,
a
Mult
i
Lay
er
Perce
ptron
(M
LP)
ne
ural
net
work
al
gorith
m
is
app
li
ed
t
o
predict
the
wind
directi
on.
The
MLP
is
a
per
ce
ptr
on
m
od
el
dev
el
oped
by
Rosenbla
tt
in
[23].
P
er
cept
ron
m
od
el
is
us
ed
to
so
lv
e
li
nea
r
pr
ob
le
m
s
on
ly
,
wh
e
reas
MLP
can
be
use
d
to
accom
plish
com
plex
probl
e
m
s
.
The
MLP
is
a
per
ce
ptr
on
that
has
t
he
ad
diti
on
al
la
ye
rs
bet
ween
the
in
put
la
ye
r
an
d
the
ou
t
pu
t
la
ye
r
know
n
as
t
he
hi
dd
e
n
la
ye
r.
The
a
rc
hitec
ture
of
M
LP
net
work
use
d
in
this
st
udy
can
be
see
n
in
Fig
ur
e
5.
T
he
in
pu
t
la
ye
r
node
s
denoted
by
X1,
X2,
an
d
X3
ob
ta
in
data
fro
m
wind
sp
ee
d
sensors
A,
B,
and
C,
respec
ti
vely
.
A
nu
m
ber
of
twenty
n
e
uro
ns i
n
the
h
i
dden
lay
er ar
e
us
e
d
t
o
im
pr
ove the
netw
ork per
f
orm
ance as a pat
te
rn ide
ntifie
r.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Wi
nd d
ire
ct
io
n sens
or
base
d on ther
m
al ane
mometer f
or
olfactory
m
ob
il
e
ro
bot
(
Hel
my W
idya
nta
r
a
)
479
Figure
5. The
a
rch
it
ect
ure
of
ne
ur
al
netw
ork
The
f
our
neur
on
s
i
n
t
he
ou
t
pu
t
la
ye
r
re
pre
sent
the
twel
ve
wind
directi
on
s
.
T
he
act
iv
at
ion
f
unct
ion
us
e
d
in
this
stud
y
is
the
bina
ry
sigm
oid
fu
nction
[23,
24]
.
Ba
ckpropa
ga
ti
on
is
a
le
arn
ing
al
go
rithm
f
or
the
arti
fici
al
neura
l
netw
ork
to
obta
in
proper
w
ei
gh
ts
betwee
n
the
ne
uro
ns
[
25
]
.
I
n
this
m
et
hod,
t
he
weig
hts
of
the
hi
dd
e
n
a
nd
outp
ut
la
ye
rs
are
up
dated
it
erati
vely
durin
g
the
le
ar
ning
proces
s
[
26]
.
T
he
back
pro
pa
gation
al
gorithm
con
s
ist
s
of
a
dvance
d
fee
d
f
orward
com
pu
ta
ti
on
a
nd
bac
k
pro
pa
gation.
T
he
fee
d
f
orwa
rd
in
hi
dd
e
n
la
ye
r
is ex
pr
es
sed by:
ij
ij
i
oj
j
x
w
w
n
et
Z
3
1
(9)
wh
e
re
i
is
the
i
-
th
node
(
i
=
1,
.
.
,3)
at
the
inp
ut
la
ye
r,
j
is
the
j
-
th
ne
uro
n
(
j
=1
,2
,
…,
20)
at
the
hid
de
n
la
ye
r,
x
ij
is
the
in
pu
t
value
of
node
i
t
o
th
e
hidde
n
ne
uro
n
j
,
w
oj
is
the
bi
as
of
t
he
hi
dden
la
ye
r,
w
ij
is
the
wei
gh
t
betwee
n
the in
pu
t
node
i
and the
hi
dd
e
n neur
on
j
.
The o
utput o
f
e
ach
neur
on is a
n
act
ivati
on
functi
on exp
ress
ed
as:
n
et
e
n
et
Z
f
Z
j
Z
j
j
1
1
)
(
(10)
The
cal
c
ulati
on
process
is the
n
a
pp
li
ed
to
th
e
outp
ut lay
er
expresse
d
as:
jk
j
k
j
Ok
k
v
z
v
n
e
t
y
1
(11)
The
val
ue
of
v
ok
is
the
bias
at
t
he
outp
ut
la
ye
r,
z
j
is
the
ou
t
put
of
each
ne
uro
n
in
the
hidde
n
la
ye
r,
and
v
jk
is
the
weig
hts
betwe
en hid
den n
e
ur
on
j
an
d
t
he out
pu
t
neur
on
k
(
k=1,2,…,
4).
k
Y
n
e
t
k
k
e
Y
n
et
f
y
1
1
)
(
(12)
The wei
ght c
orrecti
on b
et
wee
n
the
h
i
dd
e
n an
d
the
outp
ut la
ye
rs
is e
xpress
ed
as:
n
e
t
Z
f
Y
T
j
k
k
k
'
)
(
j
k
jk
z
v
(13)
wh
e
re
is
le
arn
i
ng
rate.
Me
anwhil
e,
the
weig
ht
correct
ion
betwee
n
the
in
pu
t
a
nd
t
he
hi
dd
e
n
la
ye
rs
is
expresse
d
as:
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
2
,
Fe
bru
ary 2
019
:
475
–
484
480
n
e
t
Z
f
n
e
t
j
j
j
'
i
j
ij
x
w
(14)
The u
pd
at
e
of
weig
hts a
nd b
i
as are
giv
e
n by
the
fo
ll
owin
g equ
at
io
n:
jk
jk
jk
v
v
v
ij
ij
ij
w
w
w
(15)
The
weig
hts
and
biases
gen
e
rated
in
t
he
le
arn
i
ng
process
are
the
n
us
e
d
for
the
neural
netw
ork
arc
hi
te
ct
ure
i
m
ple
m
ented
on the
m
ic
ro
cont
ro
ll
er.
3.
RESU
LT
S
A
ND AN
ALYSIS
3.1.
The
A
n
al
ys
is o
f the
Wind
Directio
n
Senso
r
Characte
rizat
io
n
a
nd
te
sti
ng
of
t
he
wind
directi
on
sen
sor
are
perform
ed
inside
t
he
wi
nd
tu
nnel
a
s
sh
ow
n
in
Fig
ure
6(
a
).
Test
s
on
the
ho
m
ogeneit
y
of
the
wind
directi
on
fo
rm
ed
in
the
tun
nel
ne
ed
to
be
accom
plished
.
The
te
st
is
c
ondu
ct
e
d
by
tur
ni
ng
the
fan
at
2.
700
rp
m
and
m
easur
i
ng
the
wi
nd
sp
ee
d
at
vari
ou
s
distances
insi
de
the
tun
ne
l.
F
igure
6(b
)
sho
ws
that
the
res
pons
e
of
wi
nd
sp
eed
in
the
wind
tunnel
is
alm
os
t
ho
m
og
e
neous
.
The
c
onfi
gurati
on
of
the
th
re
e
sens
ors
pr
es
ents
the
pe
rfo
r
m
ance
res
ult
a
s
sho
wn
in
Figure
7.
T
he
sens
or
A
has
a
good
re
spo
nse
with
the
wi
nd
c
om
ing
from
the
di
recti
on
of
0
a
nd
180
.
The
sens
or
B
has
a
good
norm
al
izati
on
sc
or
e
wit
h
the
wind
c
om
ing
f
ro
m
the
directi
on
of
120
a
nd
300
.
Me
anwhil
e,
se
ns
or
C
has
a
good
res
pons
e
wit
h
the
wind
c
om
ing
from
the
directi
on
of
60
a
nd
240
.
The
dire
ct
ivit
y
patte
rn
of
the
wind se
nsor
is
about
60
.
Figure
6.
The
c
har
act
erizat
io
n an
d
te
sti
ng
of
W
i
nd
directi
on
sen
s
or
:
(
a
) wind t
unnel,
and
(b) ho
m
og
e
neity
o
f
the
wind i
ns
ide
the t
unne
l
Figu
re
7.
The
directi
vity
patte
rn
of the
w
i
nd sen
s
or
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Wi
nd d
ire
ct
io
n sens
or
base
d on ther
m
al ane
mometer f
or
olfactory
m
ob
il
e
ro
bot
(
Hel
my W
idya
nta
r
a
)
481
At
the
wind
dir
ect
ion
wh
ic
h
ha
s
opposit
e
di
r
ect
ion
of
t
he
W
i
nd
Se
nsor
Re
v
P,
f
or
i
ns
t
ance,
t
he
wind
com
ing
from
the
directi
on
of
0
ge
ne
rates
t
he
diff
e
re
nt
no
rm
alizat
ion
sc
or
e
of
0.30
to
the
di
recti
on
of
180
.
This
di
ff
e
ren
t
r
esult
occ
ur
s
be
cause
the
posit
ion
of
t
he
the
r
m
ist
or
sign
i
ficantl
y
influ
e
nce
s
the
am
ou
nt
of
wi
nd
stream
to
reach
the
therm
ist
or
su
r
face.
I
f
th
e
wind
stream
highly
reaches
the
su
r
face,
it
will
gen
erate
a
faster
heat
release
from
the
therm
i
stor
.
The
norm
al
iz
ed
res
ponse
s
of
the
wind
se
nsor
a
re
sh
ow
n
in
Fig
ur
e
8.
This
in
dicat
es
that
the
r
esp
on
se
patte
r
n
of
t
he
sens
or
ar
ray
is
infl
uen
ce
d
by
the
ai
rf
l
ow
c
om
ing
from
diff
e
ren
t
directi
ons.
Ea
ch
wi
nd
direct
ion
will
produ
ce
a
sp
eci
fic
sens
or
patte
r
n;
there
fore,
it
can
be
us
e
d
f
or
t
he
trai
ning
proces
s of the
n
e
ural
netw
ork
t
o pr
e
dict t
he win
d d
irect
ion
.
In
th
e
trai
ning
ph
ase
,
this
ne
twork
is
fe
d
a
vecto
r
pair
consi
sti
ng
of
s
ens
or
patte
rn
s
and
ta
r
get
s
represe
nting
t
welve
wi
nd
dir
ect
ion
s.
Eac
h
directi
on
co
ns
i
sts
of
te
n
sens
or
pa
tt
ern
s
.
Th
e
nu
m
ber
of
it
erati
ons
in
this
ph
a
se
is
set
at
10
,00
0
epo
c
hs
,
as
s
how
n
in
Fig
ure
9.
The
te
sti
ng
ph
ase
of
wind
directi
on
se
nsor
is
cond
ucted
to
a
ssess
it
s
per
f
orm
ance.
Test
of
the
sensor
m
o
du
le
us
i
ng
wind
tunnel
is
perform
ed
fo
r
tw
el
ve
wind
directi
ons
with
a
sp
ee
d
of
0
.
5
-
6
m
ph
.
The
ex
pe
rim
e
ntal
resu
lt
of
th
e
wind
directi
on
sen
sor
on
the
wind
tunnel i
s
pr
ese
nted
i
n
Ta
ble
1. It s
hows
that t
he
acc
ur
acy
of
the w
i
nd se
ns
or is
91.6%.
Figure
8. The
re
spon
ses
of t
he
w
in
d
s
en
s
ors
Figure
9. The
e
rror val
ues dur
ing
t
he
trai
ning
phase
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
2
,
Fe
bru
ary 2
019
:
475
–
484
482
Table
1
.
T
he
M
easu
rem
ent R
esult
of
the
W
i
nd
D
irect
io
n
S
e
nsor
in t
he
W
in
d
T
un
nel
W
in
d
Dir
e
ctio
n
0°
30°
60°
90°
120°
150°
180°
210°
240°
270°
300°
330°
Su
ccess rate
(
%)
100
100
70
100
100
100
60
100
100
100
70
100
Av
erage of
Succes
s rate (%)
9
1
.66
3.2.
The
Perf
orm
an
ce
of th
e Olf
actory
M
ob
il
e Rob
ot E
quipped
wi
th
Wind
Directio
n S
ens
or
Test
ing
t
he
se
ns
or
m
od
ule
i
n
the
open
ai
r
beco
m
es
an
e
xciti
ng
c
halle
nge.
In
t
his
cas
e,
the
se
nsor
m
od
ule
is
i
m
plem
ented
in
an
olfacto
ry
m
ob
il
e
rob
ot
to
fin
d
the
wind
s
ource
.
Fig
ur
e
10
sh
ow
s
i
m
ple
m
entat
io
n
of
the
wi
nd
directi
on
sen
s
or
on
t
he
m
ob
il
e
ro
bot
an
d
it
s
trackin
g
al
gorithm
fo
r
1
s
econd
-
sam
pling
tim
e.
The
m
ob
il
e
robo
t
is
te
ste
d
w
it
h
sever
al
hea
ding
an
gles.
P
erfor
m
ance
te
st
on
the
m
ob
il
e
robot
in
trackin
g
the
wind
s
ource
was
accom
plished
by
blowin
g
the
wind
us
i
ng
a
n
el
ect
rical
fan
,
as
sho
wn
in
Figure
11.
T
he
r
obot
m
ov
es
wh
il
e
keep
try
ing
to
t
race
the
wind
t
ow
a
r
d
t
he
fan,
as
sho
wn
in
Fig
ur
e
12.
Th
e
exp
e
rim
ent
al
resu
lt
of
this
r
obot
is
sho
wn
in
Ta
ble
2.
T
he
su
ccess
rate
of
the
r
obot
to
f
ind
t
he
s
ource
of
wi
nd
with
head
i
ng a
ng
le
s
of
45
, 0
, a
nd
-
45
is
77.
5%, 8
7.5%,
a
nd 75%
, r
e
sp
ec
ti
vely
.
Figure
10. T
he
i
m
ple
m
entat
ion
of w
i
nd d
irec
ti
on
se
nsor:
(a) m
ob
il
e
robo
t e
qu
i
pp
e
d wit
h
t
he win
d direct
ion
sens
or
,
a
nd (b) t
he ro
bot’s tra
ckin
g
al
go
rith
m
Figure
11. Wi
nd s
ource
searc
hing
by the
olf
act
or
y m
ob
il
e r
obot
Table
2
.
T
he
S
uccess Rat
e
of
the
Ro
bot
to
Find
the
Sou
rce
of
W
i
nd
Orientatio
n
Distan
ce (
c
m
)
Av
erage
Total
25
50
75
100
Av
erage
45°
100%
80%
70%
60%
7
7
.5%
0°
100%
100%
80%
70%
8
7
.5%
80%
-
45°
100%
80%
70%
50%
7
5
.0%
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Wi
nd d
ire
ct
io
n sens
or
base
d on ther
m
al ane
mometer f
or
olfactory
m
ob
il
e
ro
bot
(
Hel
my W
idya
nta
r
a
)
483
Figure
12. T
he
robot tra
j
ect
or
y i
n
searc
h o
f win
d
s
ource
4.
CONCL
US
I
O
N
In
t
his
stu
dy,
we
hav
e
de
vel
op
e
d
a
wi
nd
di
recti
on
sens
or
us
in
g
the
rm
al
anem
o
m
et
er
pri
nciple
with
the
m
ai
n
co
m
pone
nt
of
the
po
sit
ive
te
m
per
at
ur
e
c
oeffici
ent
therm
ist
or
.
Thr
ee
anem
ome
te
rs
each
of
wh
ic
h
hav
e
dif
fer
e
nt
directi
ons
of
120
are
us
e
d
a
s
inputs
f
or
t
he
neural
net
work
t
o
determ
ine
the
directi
on
of
t
he
wind
aut
om
at
i
cal
ly
.
The
ex
pe
rim
ental
resu
lt
s
sh
ow
that
t
he
wind
se
ns
or
syst
e
m
is
able
to
rec
ognize
t
welve
wind
directi
on
s
with
the
accu
racy
of
91.
6%
.
A
n
olfa
ct
or
y
m
ob
il
e
ro
bo
t
e
qu
i
pp
e
d
with
t
his
sens
or
syst
e
m
ca
n
nav
i
gate to
a
w
ind
source
in t
he op
e
n
ai
r
with a
su
cces
s r
at
e of
80
%
.
ACKN
OWLE
DGE
MENTS
This
researc
h
was
ca
rr
ie
d
ou
t
with
fina
ncia
l
ai
d
s
upport
f
ro
m
Lem
bag
a
Penelit
ia
n
dan
Pengab
dian
Kep
a
da
Ma
sya
rak
at
(L
PPM
)
Insti
tut
Teknolo
gi
Sepulu
h
Nopem
ber
(ITS)
S
ur
a
baya,
and
the
Mi
ni
stry
of
Re
search
, Tec
hnology a
nd Hi
gh
e
r
E
du
cat
i
on
of the Re
publ
ic
o
f Ind
on
e
sia
(Kem
enr
ist
ekd
ikti
RI
).
REFERE
NCE
S
[1]
Babu
SD
,
Acha
rj
y
a
AS
,
Alam
S
.
Design
And
Constr
uct
ion
of
Mic
rocontroller
Based
Wind
S
pe
ed
And
Direc
t
io
n
Monit
oring S
yst
em
.
In
Strategi
c Te
chno
log
y
.
Co
x'
s Ba
za
r
.
2014
: 284
-
288.
[2]
Kim
KH
,
Le
e
SB
,
Dae
kwangW
oo
,
Bae
GN
.
Infl
uenc
e
of
W
ind
Dire
ction
And
Speed
on
The
T
ra
nsport
of
Parti
c
l
e
-
bound
PA
Hs
in
a
Roadwa
y
Env
i
ronm
ent
.
At
mos
pheric
Pol
lu
ti
on
Re
search
.
2015;
6(6):
1024
-
1034
.
[3]
Li
JG
,
Yang
J,
Cui
SG
,
Geng
LH
.
Speed
Li
m
it
a
tion
of
a
Mobile
Robot
and
Meth
odolog
y
of
Tr
ac
i
ng
Odor
Plu
m
e
in
Airflow
Envi
ron
m
ent
s.
Jounal E
l
sev
ie
r
Proce
d
ia Engi
ne
ering
.
20
11;
15:
1014
-
10
45.
[4]
Akin
y
emi
ML
,
Emete
re
ME,
A
kinwum
i
SA
.
Dy
nami
cs
of
W
ind
Strengt
h
and
W
ind
Dire
ction
on
Air
Pollut
ion
Dispersion.
Asia
n
Journal
o
f App
li
ed
Scienc
es
.
20
16;
4(2):
422
-
42
9.
[5]
Rahi
m
i
MR,
Haji
gh
ase
m
i
S,
Sanae
i
D.
Desi
gning
and
Sim
ula
ti
on
for
V
er
ti
c
al
Moving
C
ontrol
of
UA
V.
Inte
rnational
Jo
urnal
of El
e
ct
ri
c
al
and
Comput
er
Engi
n
ee
ring
(
IJE
CE)
.
2013;
3(5
):
651
-
659.
[6]
M
Riva
i,
Rend
yans
y
ah
,
D
Purw
a
nto.
Impl
eme
ntation
of
Fuzzy
Logic
Control
In
Ro
bot
Arm
For
Sea
rching
Locati
o
n
of
Gas
Leak
.
Int
ern
ational Sem
ina
r
on
Intelli
g
en
t
Te
chno
log
y
a
nd
Its
Applicati
on
.
Suraba
y
a. 2015:
69
-
74.
[7]
K.
Harri
ngton
,
N.
Daniel,
R.
Gurka,
A.
Hara
ri,
G.
Z.
A
.
L
iberzon.
Moth
-
Insp
ire
d
Nav
iga
t
ion
Algorit
hm
in
a
Turbul
en
t
Odor
Plum
e
From
a
Puls
at
ing
Source
.
PLOS ONE
,
201
8;
13(6):
1
-
18.
[8]
Abus
al
oua
A,
Sheli
q
R.
Dispersio
n
Models
of
At
m
ospheric
Air
P
oll
uta
n
ts.
Uni
ve
r
sity
Bul
l
et
in
.
201
6;
18(3):
56
-
79.
[9]
Zha
ng
T,
Li
H
,
W
ang
S.
Inv
erse
l
y
Tracki
ng
Indoor
Airbor
ne
Parti
cles
to
Loc
ate
The
ir
Rel
e
ase
Source
s.
At
mos
pheric
En
vi
ronm
ent
.
2012
;
55(3):
328
-
338.
[10]
Lu
TF.
Indoor
Odour
Source
Loc
alisati
on
Us
ing
Robot:
Init
ial
Loc
a
ti
on
and
Surge
Distanc
e
Ma
tt
er
.
Robot
i
cs
and
Aut
onomous
Sys
te
ms
.
2013;
61:
637
-
647.
[11]
Voges
N,
Chaff
i
ol
A,
Lu
ca
s
P,
Marti
ne
z
D.
Re
a
ct
iv
e
Sear
chi
ng
and
Infota
x
is
in
Odor
Source
Lo
ca
l
iz
a
ti
on
.
PLO
S
Computati
onal Bi
ology
.
2014
;
1
0(10):
1
-
14.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
13
, N
o.
2
,
Fe
bru
ary 2
019
:
475
–
484
484
[12]
Zha
ng
J,
Gong
D,
Zha
ng
Y.
A
Nichi
ng
PS
O
Ba
sed
Multi
Robot
Coopera
ti
on
M
et
hod
for
Loc
a
lizing
Odor
Source
s.
Neurocomputi
ng
.
2013;
123:
308
-
317.
[13]
Jatmiko
W
,
Jova
n
F.
Robots
Im
ple
m
ent
a
ti
on
For
Odor
Source
Lo
ca
l
iz
a
ti
o
n
Us
ing
PS
O
Algorit
hm
.
World
Sci
en
ti
f
ic
and
Engi
n
ee
ring
Ac
ad
emy
and
S
oci
e
ty
.
2011;
4(1
0):
115
-
125.
[14]
Ta
n
Y,
Zh
eng Z. Re
sea
r
ch
Adva
nce
in
Sw
arm R
oboti
cs.
Def
enc
e
Technol
og
y
.
20
13;
9(1):
18
-
39.
[15]
Chen
B,
Zhu
YQ
,
Yi
Z,
Qin
M,
Huang
QA
.
Tem
per
at
ure
Eff
ects
on
the
W
ind
Dire
ction.
Senso
rs
.
2015;
15(12):
29871
–
29881.
[16]
Hultmark
M,
Sm
it
s
AJ
.
Te
m
p
era
tur
e
Corre
c
tions
for
Constant
Te
m
per
a
ture
and
Constant
Curre
nt
Hot
W
ire
Anem
om
et
ers.
Me
asur
eme
nt
S
c
ie
nc
e
And
Techn
ology
.
2010;
21: 1
-
4.
[17]
Zhu
Y,
Chen
B
,
Qin
M,
Huang
QA
.
2
D
Micromac
hine
d
Th
er
m
al
W
ind
Sensors
A
Revi
ew.
I
EE
E
In
te
rnet
o
f
Things
Journal
.
2014;
1(3):
216
-
232.
[18]
Kham
shah
N,
A
bdal
l
a
AN
,
Koh
SP
,
Rashag
HF
.
Iss
ues
and
Te
mpe
rat
ur
e
Com
pensa
ti
on
Techni
q
ues
for
Hot
W
ire
The
rm
al
Flow
Senso
r:
A Review
.
Int
ernati
onal
J
ournal
of the
Ph
ysic
al
Scienc
es
.
2011;
6(14):
327
0
-
3278.
[19]
Kam
iński
Z,
Cz
y
ż
Z.
The
Ch
aract
er
isti
cs
of
Th
e
Opera
ti
ng
Par
a
m
et
ers
of
The
V
ert
i
ca
l
Axis
W
in
d
Turbi
ne
For
T
he
Sele
c
te
d
W
ind
Speed.
Ad
vances in Sc
i
ence
and
Te
chnol
ogy
Re
s
ea
rch
Journal
.
2017
;
11(1):
58
-
65.
[20]
Yong
TH,
Dol
SS
.
Design
and
D
eve
lopment
o
f
L
ow
-
Cost
W
ind
Tunne
l
for
Educ
a
t
iona
l
Purpos
e.
M
ate
rials
Sc
ie
nc
e
and
Engi
n
ee
ring
.
2015;
78:
1
-
11.
[21]
Garve
y
L,
Hass
an
L,
Nicho
ls
K,
Paquin
A.
Equ
ine
Lung
Func
t
i
on
Testing
Dev
i
ce
.
In
No
rth
ea
st
Bioe
ngineeri
ng
Confer
ence. Phil
ade
lph
ia
.
2017.
[22]
Prohask
y
D,
W
at
kins
S.
Low
C
ost
Hot
-
element
Ane
mom
et
ry
V
erses
the
TFI
C
obra
.
In
19th
A
ustral
asi
an
Fluid
M
ec
hanics Conf
ere
nc
e. Me
lbour
ne.
2014
:
260
-
26
3.
[23]
Pancha
l
G
,
Gan
at
ra
A
,
Kos
ta
Y
P,
Panchal
D.
B
eha
viour
An
aly
s
is
of
Multi
l
a
y
er
Perc
ept
rons
wi
th
Multi
pl
e
Hidde
n
Neurons a
nd
Hi
dden
L
a
y
ers.
Int
ernati
onal
Journ
al
of
Computer
Theory
and
Eng
i
nee
ring
.
2011;
3
(2):
332
-
337
.
[24]
M Riva
i,
F Budi
m
an,
D Purwanto,
J Sim
amor
a.
Mea
t
Freshness
Ide
nti
f
ic
a
ti
on
S
y
stem Us
ing
Gas
Sensor Arra
y
And
Color
Sensor
In
Conjunc
t
ion
W
it
h
Neura
l
Network
Pat
te
rn
Rec
ognition
.
Jo
urnal
of
Theore
ti
cal
and
Appl
i
e
d
Information
Tec
hnology
.
2018,
9
6
(12):
3861
-
387
2.
[25]
M
Riva
i,
A
Arif
in,
EI
Agus
ti
n.
Mix
e
d
Vapour
I
dent
ification
Us
i
ng
Parti
ti
on
Col
umn
-
QCMs
And
Arti
ficia
l
Neura
l
Net
work
.
Inte
r
nat
ion
al
Conf
er
enc
e
on
Infor
m
at
ion,
Com
m
unic
a
ti
on
T
ec
h
nolog
y
and
S
y
stem.
Sur
abaya.
2016:
172
-
177.
[26]
Rao
CS
.
Design
of
Artif
ic
i
al
Int
el
li
g
ent
Controll
er
for
Autom
at
i
c
Gen
er
ation
Co
ntrol
of
Two
Ar
ea
H
y
droth
erma
l
S
y
stem.
In
te
rnat
ional
Journal
of
El
e
ct
rica
l
and
C
omputer
Engi
n
e
ering
.
2012
;
2(2)
:
183
-
190.
Evaluation Warning : The document was created with Spire.PDF for Python.