Indonesian J
ournal of Ele
c
trical Engin
eering and
Computer Sci
e
nce
Vol. 2, No. 2,
May 2016, pp
. 380 ~ 389
DOI: 10.115
9
1
/ijeecs.v2.i2.pp38
0-3
8
9
380
Re
cei
v
ed Fe
brua
ry 20, 20
16; Re
vised
Ap
ril 10, 201
6; Acce
pted
April 27, 201
6
Lifetime Prolonging for Clustered Heterogeneous
Wireless Sensor Networks by SEP-FUZZY
Basim Ab
oo
d
1
, Aliaa Hussien
2
, Muha
mmed Shaemy
a
l Nisar
3
, Yu
Li*
4
1,3,
4
Nationa
l La
borator
y for Op
toelectro
n
ics, Huaz
hon
g
Un
i
v
ersit
y
of Sci
e
n
c
e and T
e
chno
log
y
W
u
h
an,
Chin
a 43
00
74
2
Baghd
ad U
n
iv
ersit
y
, Co
ll
ege
of Science,
De
partment of astronom
y Ba
gh
d
ad, Iraq
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: bas_e
ng
84@
yma
il.com
A
b
st
r
a
ct
T
he most i
m
p
o
rtant cons
ide
r
ation i
n
d
e
sig
n
in
g
protoc
ols
for w
i
reless sensor
netw
o
rks is t
h
e
ener
gy
co
nstraint
of no
des beca
u
se in most
cases
battery rech
argi
ng
is inc
onv
eni
e
n
t or i
m
possi
b
l
e
.
T
herefore, ma
ny
res
earch
es
hav
e be
en d
one
to over
c
o
me
this
d
e
m
er
it. Cluster
ing
i
s
on
e
of the
ma
in
appr
oach
e
s i
n
desi
gni
ng sc
al
abl
e a
nd
en
er
gy-efficie
n
t pro
t
ocols for W
S
Ns.In this p
a
p
e
r, w
e
pro
pos
ed
a
novel schem
e
to inv
e
stigate the cluster, the fu
z
z
y
lo
gic c
l
uster stable
election
protoc
ol (SEP - FUZZY)
,
w
h
ich uses fu
z
z
y
lo
gic infer
e
n
c
e system (FIS) in the
cluster process. We compar
e our te
chni
que w
i
th tw
o
appr
oach
e
s (L
EACH, and SE
P) to show
that using a
mu
lti
para
m
eter F
I
S enha
nc
es th
e netw
o
rk lifeti
m
e
signific
antly. Si
mu
lati
on res
u
lt
s de
monstr
ate
that t
he netw
o
rk lifeti
m
e
ac
hiev
ed by th
e prop
osed
meth
od
coul
d b
e
incre
a
sed
by
ne
arly
27
%
more th
a
n
that
obta
i
n
e
d
by LEACH
pr
otocol, an
d by near
ly
2
3
%
mo
r
e
than that obta
i
ned by Sta
b
le Electio
n
Protoc
ol.
Ke
y
w
ords
:
WSNs, clustering,
fuz
z
y logic, SEP
Copy
right
©
2016 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
A Wirele
ss S
ensor Netwo
r
k (WS
N) i
s
a comb
i
nation of
au-tonom
o
u
s
dist
ribute
d
sensors
to monito
r e
n
vironm
ental
of phy
sical
co
ndition
s
and to
pa
ss the a
ggreg
ated
sen
s
e
d
data
throug
h the n
e
twork to a m
a
in ba
se stati
on or si
nk.
In WSN,
ene
rgy efficie
n
t
comm
uni
cati
on i
s
the
su
b
j
ect of
su
rvival of a
net
wo
rk. A
s
a
result, the
re
sea
r
che
r
s are mo
stly fo-cuse
d
to
ward
s e
n
e
r
gy effi
cient
co
mmu
nicatio
n
, en
e
r
gy
manag
e-m
e
n
t, and extend
ing the n
e
two
r
k lifetime. M
o
st of the tim
e
, study on
Wirel
e
ss Se
n
s
or
Networks ha
s a
s
sum
ed
h
o
moge
neo
us nod
es. B
u
t i
n
reality, ho
mogen
eou
s
node
s
also h
a
ve
different leve
ls of initial energy and
depletio
n/
drai
n rate. This leads to th
e resea
r
ch on
hetero
gen
eo
us n
e
two
r
ks
whe
r
e two o
r
mo
re type
s of node
s a
r
e co
nsi
dered
and the
mo
re
powerful sen
s
or n
ode
s a
c
t as clu
s
te
r he
ads a
nd ha
nd
le resou
r
ce-d
emandi
ng tasks.
A Sensor No
de i
s
co
nsi
s
ti
ng of
sen
s
ors, p
r
o
c
e
s
sor,
tran
sceiver
and
po
we
r u
n
its a
s
sho
w
in
g in F
i
gure
1. As
we
kno
w
tha
t
while
d
e
-si
gning
a ro
uting protocol
a se
nsor n
o
de is
limited ene
rg
y supply, so
available en
e
r
gy at t
hat nodes mu
st be
a major
con
s
t
r
aint. For e
n
e
r
gy
efficien
cy, extensibility of l
i
fetime, scala
b
ility
and perfor-ma
n
ce, cl
uster
ba
se
d routing proto
c
ol
enforce
s a structure out of di
fferent ro
uting proto
c
ol
s.
In clu
s
te
r b
a
s
ed
routing
proto
c
ol
s,
se
nso
r
s a
r
e
di
vided into
di
fferent
clu
s
te
rs after
sele
cting so
me node
s a
s
clu
s
ter h
e
ads am
ong t
hem, so that
sen
s
or n
o
d
e
s commu
nicate
informatio
n o
n
ly to cluste
r head
s and
aggregate inf
o
rmatio
n to base
station. Clu
s
terin
g
is
an
efficient way
to redu
ce e
n
e
rgy con
s
um
ption and ext
end the life time of the net
work, doi
ng d
a
ta
aggregatio
n
and fu
sion in
orde
r to red
u
ce th
e num
ber of tra
n
smitted messa
ges to the
b
a
se
station [1].
The rest of t
h
is p
ape
r is
orga
nized a
s
fo
llows. Rel
a
ted work (p
rior
art
s
) a
n
d
related
con
c
e
p
ts
of
desi
gning
the
WS
Ns
and
apply-in
g the
routin
g al
go
rithms to ext
end th
e n
e
twork
lifetime are pre
-
sented in
part 2. In
Part 3,
the
pape
r pre
s
e
n
ts the routi
ng model for the
prop
osed rou
t
ing
metho
d
. Perform
a
n
c
e evaluation
i
s
prop
osed i
n
Part 4. Fin
a
ll
y, con
c
lu
sion
is
pre
s
ente
d
in Part 5.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 2, May 2016 : 380 –
389
381
Figure 1. A typical
Wirel
e
ss Sensor
Net
w
ork
2. Related Work
The re
sea
r
ch
es of
wi
rele
ss
se
nsor network
cluste
rin
g
algo
rithm
are of g
r
eat
im
portan
c
e
to improve th
e netwo
rk performan
ce a
n
d
also of
g
r
eat
importa
nce to pra
c
tical a
p
p
lication. Th
ere
are several propo
sed
clu
s
tering al
go
rithms in the literature.
There a
r
e
two
cate
gori
e
s of
clu
s
te
ri
ng
scheme
s
,
those a
ppli
ed in
ho
mo
gene
ou
s
netwo
rks, referred to a
s
h
o
moge
neo
us
clu
s
teri
n
g
sch
e
mes,
and th
ose
appli
ed i
n
heteroge
ne
ou
s
netwo
rks,
ref
e
rred to
a
s
h
e
terog
ene
ou
s clu
s
te
ring scheme
s
.
Th
e most popul
ar
homo
gen
eo
us
clu
s
terin
g
alg
o
rithm
s
in
clu
de lo
w- e
n
e
r
gy adaptive
clusteri
ng hi
erarchy (LEA
CH) [2] a
nd p
o
wer-
efficient gath
e
ring i
n
se
nsor info
rmatio
n syst
em
s (PEGASIS) [3]. On the other ha
nd, hyb
r
id
energy-
effici
ent di
stribut
e
d
cl
us
te
ring
(HEED)
[4], and dist
ribut
ed
e
nergy
ef
ficient clu
s
tering
(DEEC) [5] are hetero
gen
e
ous
clu
s
terin
g
algorith
m
s.
The LEACH proto
c
ol sel
e
cts
CHs pe
riodi
cally
and
drain
s
en
ergy uniformly
by role
rotation. Ea
ch nod
e ma
kes a
de
cisi
o
n
wh
ether
or not to be
a
CH
acco
rdi
ng to a
unifo
rmly
distrib
u
ted
probability. In
PEGASIS, node
s will
be
orga
nized to
form a
chai
n, whi
c
h
ca
n
be
comp
uted by
each no
de
or by t
he b
a
s
e
station. T
he re
qui
reme
nt
of global
knowl
edge
of
the
netwo
rk to
pol
ogy makes t
h
is meth
od d
i
fficult to implement. HEE
D
is
a dist
rib
u
ted cl
uste
rin
g
algorithm, which
selects the CHs
stochasti
c
ally
. The election probability of each node is
correl
ative to the remai
n
i
ng ene
rgy. In het
eroge
ne
ous e
n
viro
n
m
ents, the l
o
w-ene
rgy n
ode
s
coul
d o
w
n la
rger
electio
n
p
r
oba
bility than the hig
h
en
ergy no
de
s in
HEED. Th
e
hetero
gen
eity of
node
s in te
rm
s of thei
r en
ergy is
con
s
ide
r
ed in
DEEC.
DEEC h
a
s th
e merit
of bei
ng a di
stri
but
ed
clu
s
terin
g
al
gorithm.
Ho
wever, the
perfo
rm
an
ce
of homog
e
neou
s
sche
mes i
s
po
or for
hetero
gen
eo
us networks becau
se
th
e
low-e
ner
gy
node
s
co
uld
have
a
hig
her proba
bility of
electio
n
than
the high en
ergy node
s [6].
Stable ele
c
tio
n
protocol
(S
EP) studi
es t
he impa
ct of
hetero
gen
eity, in terms of e
nergy of
the nodes.
To elect the
CHs, SEP uses a
wei
ght
ed probability method based on remai
n
ing
energy in th
e
nod
es. T
h
is
coul
d p
r
olo
n
g
the
stab
ility
perio
d of th
e
netwo
rks
(sta
bility is d
e
fin
e
d
as
the time f
r
om the beginni
ng of the
network
P
r
ocess
until
the
firs
t node dies
). In SEP a
n
adju
s
table p
e
r
ce
ntage
of the no
des
ha
ve highe
r en
ergy than th
e
other n
ode
s.
Acco
rdi
ngly, a
modified p
r
o
bability is de
fined to con
s
ider the
re
si
dual e
nergy of the nod
es. Based o
n
this
probability, the length of
used
epoch in
LEACH is in
creased.
The
authors’
show that
compared
to LEACH, SEP c
an increas
e
the
s
t
ability period of the network
.
Bala et al [7] c
o
ns
idered
determinis
t
ic
-SEP
(D-SEP
), for
elec
ting c
l
us
ter heads
in a
distrib
u
ted
st
yle in two, t
h
ree,
and
m
u
lti-leve
l hie
r
archical
wirel
e
ss sen
s
o
r
netwo
rks. Th
e
s
i
gnific
ant improv
ement has
been us
ing D-SEP
in c
o
mparis
on with SEP in
terms
of energy
consum
ption,
data transmi
ssi
on and network lifetim
e
to Base
stati
on. D-SEP protocol
goal i
s
to
enlarge the lifetime and stability of the network i
n
the
presence of
heter
ogeneous n
odes. Si
nce
clu
s
ter he
ad
s consume m
o
re en
ergy th
an clu
s
te
r m
e
mbe
r
s in re
ceiving an
d sensi
ng data from
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Lifetim
e
Prolonging for
Clustere
d Heterogeneous WS
Ns by SEP-F
U
ZZY
(Yu Li)
382
their comp
on
ent
nod
es, pe
rformin
g
sign
al
dispen
satio
n
and
tran
sfe
r
the
agg
reg
a
t
ed data to
n
e
xt
node o
r
ba
se
station. D-S
EP network li
fetime is
increased by 4.4
times over S
EP. The resu
lts
clearly show t
hat the
stabili
ty period of
D-SEP is
l
onger as
com
pared to
SEP. The authors in [8]
proposed an
enhanced SEP (E-SEP), it
is three ti
re node protocols. E-SEP operates in a WS
N
under three-tier energy het
er
ogeneity. E-SEP introduces addi
tional nodes
nam
ed intermedi
ate
node
s th
at a
c
t a
s
a
bri
d
ge b
e
twee
n
the no
rmal
a
nd a
d
vanced
nod
es. T
h
e
ene
rgy of t
h
e
interme
d
iate
node
s i
s
set
betwe
en th
at
of no
rmal
an
d adva
n
ce n
o
des.
The
work in
[9] p
r
op
o
s
ed
Dis
t
ance based –SEP (DB-SEP), the DB-SEP has
two part
s
, firs
t on the nodes
initial energy
and
se
con
d
ly the sel
e
ctio
n
of CH
is
dist
ance ba
se
d, the one
clo
s
e
to the si
nk
will have
hig
her
cha
n
ce to
be
come
CH Clu
s
terin
g
hi
erarchy i
s
u
s
e
d
t
o
e
s
tabli
s
h th
e net
work, i
n
orde
r to
redu
ce
the correl
ated
data, the CH performs fu
si
on functio
n
.
In this work
,
we
propos
e
a new protoc
ol
SEP-Fuz
z
y
improves
SEP
protocol us
ing Fuzz
y
Logi
c. In SEP, the election probabilities of cl
uste
r head are
weighted by t
he i
n
itial energy
of a
node
rel
a
tive to that of ot
her
nodes in t
he net
work.
SEP-FUZ
Z
Y prov
ides a l
onger
stabilit
y
perio
d a
nd a
lowe
r in
stabili
ty period
and
increa
se
s lif
e time of n
o
d
e
s.
We
study
the effect
of
our
SEP-FUZZY
protocol to heterogeneity
parameters
capturing energy imbal
ance in the network.
The inp
u
ts th
at we con
s
id
er in the fu
zzy syst
em are
:
Remaini
ng
Energy, No
rmalize
Di
stan
ce
from
sen
s
o
r
n
ode to
cl
uste
r hea
d (nod
e-CH), an
d
No
rmalize
Di
stan
ce from
clu
s
t
e
r h
ead
to Ba
se
station
(CH-B
S
). The
s
e
pa
ramete
rs a
r
e
not
so
cl
o
s
ely r
e
la
te
d and
ca
n
e
a
s
ily
w
o
rk w
i
th th
es
e
hetero
gen
eo
us
paramete
r
s by
usi
ng f
u
zzy logi
c.
A
l
so
a fu
zzy
system
doe
s not n
eed
m
u
ch
comp
utationa
l complexity; con
s
e
que
ntly it is su
itable for WSN. Accordin
g to wha
t
was said, we
prop
osed a
di
stribute
d
met
hod a
nd e
a
ch
node it
se
lf m
a
ke
s d
e
ci
sio
n
about b
e
ing
clu
s
ter
head
or
not. This met
hod m
u
st
wo
rk in
all e
n
viro
nments an
d
so do
esn’t
nee
d no
des coo
r
dinate
s
. In thi
s
method
by
choo
sing
suitable i
nput
s f
o
r fu
zzy sy
stem, is mo
re
efficient
tha
n
the
existe
nce
method an
d b
e
tter clu
s
ter
will be mad
e
.
3. Stable Election Proto
c
ol and Fuzz
y
Approa
ch
3.1. Stable Election Protocol (SEP)
A. Net
w
o
r
k Model
In this section, we describ
e the SEP protocol. Assume that
there are N sensor nodes.
Nod
e
s al
way
s
have data to transmit to
a base
statio
n, which is often far away from the se
nsi
n
g
area. T
he n
e
twork i
s
o
r
gani
zed into
a clu
s
tered
hiera
r
chy where
every
cluster
ha
s a
CH,
respon
sibl
e f
o
r exe
c
utin
g
fusion
fun
c
tion to
r
edu
ce
co
rrel
a
ted
d
a
ta produ
ce
d
by the
sen
s
or
node
s within
the same cl
uster. Th
e CHs di
re
ctly transmit the a
ggre
gated d
a
t
a to the base
station. We
suppo
se that
the nod
es a
r
e
stationa
ry.
SEP does
not requi
re
energy knowledge shar
i
ng but is based on assigni
ng
weighted
election probabilities
of each n
ode to becom
e
a
CH according to
their
respect
i
ve energy. By
using thi
s
approach,
SEP ensures
that
the CH
i
s
randomly sele
cted based on the fraction
of
energy of each node. Thi
s
also results in
a uni
form di
stribution of en
ergy co
nsum
ption.
In SEP, the election probabilitie
s are
weighted by
the initial
en
ergy
of a node rel
a
tive to
that of oth
e
r
node
s i
n
the
netwo
rk.
Thi
s
prolong
s
the
time inte
rval
befo
r
e th
e d
eath of t
he first
node (stabilit
y
perio
d), which
i
s
cruci
a
l
for
m
any
appli
c
ation
s
whe
r
e th
e fe
edba
ck from
the
sen
s
o
r
netwo
rk mu
st be rel
i
able.
Let
E
be the
initial energy of
basi
c
se
nsors, and
m
be the
fraction of CHs, whi
c
h own
times mo
re e
nergy than th
e norm
a
l one
s. Thu
s
, there are
to CHs equi
ppe
d with an initial
energy of
1
α
E
; and
1
m
N
, (basi
c
se
nso
r
s)
with a
n
initial ene
rg
y of
. Thus
, the total
initial energy of the two leve
l hetero
gen
e
ous n
e
two
r
ks is:
E
N
1–
m
E
m
N
E
1
α
NE
1
α
m
(1)
So, the total
energy of the syst
em i
s
increased by a factor of
1
α
m
.
Let
P
be the weig
hted ele
c
tion proba
bil
i
ty
of advance node
s.Opti
mum pro
babi
lity
P
of
each nod
e to become CH can be calcula
t
ed by (2).
P
α
1
α
(2)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 2, May 2016 : 380 –
389
383
The thre
sh
old
is given by (3).
T
s
.
ifs
∈
G
0
o
therwise
(3)
In this pape
r,
we co
nsi
der a sen
s
o
r
ne
twork con
s
ist
i
ng of N se
n
s
or
node
s de
ployed
over a va
st
field to co
ntinuou
sly m
onitor the
e
n
vironm
ent. For a
WS
N we m
a
ke some
assumptio
n
s
about the sen
s
or n
ode
s an
d the unde
rlying network m
odel.
Senso
r
no
de
s are d
eploye
d
rand
omly.
All sensor n
o
des a
nd the b
a
se
station
are stationa
ry after
the deplo
y
ment phase.
Nod
e
s are
capabl
e of a
d
j
u
sting
the tra
n
smi
ssi
on
po
wer a
c
cordin
g to the
dist
ance of the
receiver n
ode
s.
The
dista
n
ce
between
no
des can
be
comp
uted
ba
sed
on
the
receive
d
sign
al st
ren
g
th.
Therefore, th
ere is n
o
nee
d for se
nso
r
n
ode
s to kno
w
their exact lo
cation
s.
All sensor n
o
des h
a
ve not the same a
m
ount
of energ
y
when they are initially de
ployed.
The ba
se stat
ion nee
d not be located far away from th
e sen
s
in
g reg
i
on.
B. Energ
y
C
onsumptio
n Model
For the
re
ali
s
tic, the first
ord
e
r
radi
o
m
odel th
at will be
used
in LEACH [
2
], as a
comm
uni
cati
on model be
tween senso
r
node
s. Both the free space
d
power loss) an
d the
multipath fadi
ng (
d
po
we
r lo
ss)
cha
nnel
model
s a
r
e
u
s
ed, d
epe
ndi
ng on
the di
stance
betwee
n
the transmitter and receiver. The ene
rgy c
onsumpti
on in transmi
tting a packe
t with
k
-bit
s over
distan
ce
d
.
E
is t
he am
ount of
ene
rgy con
s
umption
per
bit to run
the
tran
smitter o
r
re
ceive
r
cir
c
uit
r
y
.
E
, and
E
is the amount of energy
per bit dissip
a
ted in the RF amplifier according to
the distan
ce
d
whi
c
h can be
obtaine
d from
(4), and
(5) a
s
belo
w
.
E
T
k
k
E
E
d
,
ifd
d
k
E
E
d
,
ifd
d
(4)
d
(5)
The amo
unt of energy con
s
umptio
n in receivin
g a pa
cket with
k
bits can be
cal
c
ul
ated by (6).
E
R
k
k
E
(6)
The ra
dio en
ergy mod
e
l p
a
ram
e
ters prese
n
t details i
n
Figure 2, and Table 1.
Figure 2. First order
Radi
o model
Transm
it
E
l
ectron
i
cs
Tx Am
plifie
r
Recei
v
e
E
l
ectron
i
cs
K b
i
t p
a
ck
et
K b
i
t p
a
ck
et
d
E
n
T
(
k
)
E
n
T
(
k
)
E
elec
* k
E
am
p
*
k
*
d
n
E
n
R
(
k
)
E
elec
* k
E
elec
* k
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Lifetim
e
Prolonging for
Clustered
Heterogeneous WS
Ns by SEP-F
U
ZZY
(Yu Li)
384
Table 1. Para
meters of the first radi
omod
el
Parameter value
50
10
⁄
⁄
100
⁄
⁄
3.2. Fuzzy
Clusterin
g
Ap
proach
The
con
c
e
p
t of fuzzy log
i
c was intro
d
u
ce
d by Za
d
eh in th
e mi
d-19
60
s [10]
as
an
extensio
n of the co
ncept
of an
ordina
ry fuzzy set. Since the
n
, its appli
c
ation
s
have
rapi
dl
y
expand
ed in adaptive co
ntrol syste
m
s
a
nd system id
entification. It
has the adv
antage
s
of easy
impleme
n
tation, robu
stne
ss, and ability to approximat
e to any nonli
near m
appi
ng
[11].
In Fuzzy
Clu
s
tering m
odel,
whe
n
a
ba
sic sen
s
o
r
d
e
tects an
event a
nd wants to transmit
it
s pac
ket
s
,
it sele
ct
s a be
st
clu
s
t
e
r he
ad.
To
achi
e
v
e this, we make u
s
e of
fuzzy logic.
The
obje
c
tive of fuzzy logic is
therefo
r
e to calcul
ate
the optimal value of the best cl
uster h
e
ad th
at
depe
nded
wh
ich d
epe
nd
s
on the
remai
n
ing e
nergy
RE, D (nod
e-CH), and
D
(CH-BS) a
s
shown
in figure
s
. (3)
and (4
) re
sp
e
c
tively.
Figure 3 and
figure.4, sho
w
s t
he fu
zzy
logic
with three input vari
able
s
(RE, , D (no
d
e
-
CH), and D (CH-BS)), an
d
an output (F
i
t
ness value),
with universal
of discou
r
se
[0. .
.5],
[0. . .1],
[0. .
.1], and
[0. .
.1], resp
ective
ly. Fuzzy Clu
s
terin
g
use
s
five membe
r
ship fu
nction
s for e
a
ch
input and a
n
output variabl
e, as sh
own in Figure 5.
Figure 3. Fuzzy stru
cture with three in
p
u
ts RE,
D(n
ode
-CH), D(CH-BS), a
nd one o
u
tput
of
Fitness Value
of CH
Figure 4. Fuzzy Clu
s
terin
g
System
Arc
h
itec
ture
In Fuzzy Cl
u
s
terin
g
, the f
u
zzified valu
e
s
a
r
e p
r
o
c
e
s
sed
by the in
feren
c
e
engin
e
, whi
c
h
con
s
i
s
ts
of a
rule
ba
se
an
d vari
ou
s me
thods to infe
r the
rule
s. T
a
bles 2–
6
sho
w
the
IF-T
HE
N
rule
s u
s
e
d
in
Fuzzy Cl
ust
e
ring,
with a
total numb
e
r
of
5
125
for
the fuz
z
y
rule bas
e
.
As
example, IF
RE is Very
High
an
d
Distance
to
CH
is Nea
r
a
nd Dista
n
c
e
to BS
is
Ve
ry
Nea
r
THEN Fitn
ess value is Ve
ry Good.
All these rul
e
s a
r
e p
r
o
c
e
s
sed in a p
a
rallel
mann
er
by a fuzzy inf
e
ren
c
e
engin
e. At the
end,
the def
uzzificatio
n
finds
a singl
e cri
s
p output
value from
th
e solution fu
zzy
sp
ace. T
h
is
value re
presents the
nod
e co
st. Pra
c
tice d
e
fuzzif
ication is
ca
rri
ed out u
s
ing c
e
n
t
er
o
f
gra
v
ity
method given
by (7) [12].
∗
(7)
Whe
r
e
is
the output of ru
le base
, and
is the center of the
output membershi
p
function for
rule ba
se num
ber.
Rule
Base
Infer
e
nc
e
En
gi
ne
D
e
f
u
z
z
i
f
i
c
a
t
i
o
n
Fu
z
z
y
Fu
z
z
y
Outputs
Outputs
Fu
z
z
y
Fu
z
z
y
I
nputs
I
nputs
F
itne
ss Va
lue
F
itne
ss Va
lue
Re
ma
ining
Energ
y
Re
ma
ining
Energ
y
f
u
z
z
i
f
i
c
a
t
i
o
n
Distance (
node
–
CH
)
Di
st
ance (
node
–
CH
)
Di
st
ance (
C
H
–
BS
)
Di
st
ance (
CH
–
BS
)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 2, May 2016 : 380 –
389
385
Figure 5. Membershi
p
gra
ph for thre
e inputs (r
em
ain
i
ng ene
rgy (RE), normali
ze
distan
ce (n
o
de-
CH), and no
rmalize
Dista
n
c
e (CH-BS)
a
nd the output
(fitness value
)
.
Table 2. IF-T
HEN rule
s, where e
n
e
r
gy is very low
D(CH
)
D (BS)
V. Near
Near
Medium
Far
V. Far
V. Near
Near
Medium
Far
V. Far
Normal
Normal
Bad
Bad
V. Bad
Bad
Bad
V. Bad
V. Bad
V. Bad
Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
Table 3. IF-T
HEN rule
s, where e
n
e
r
gy is low
D(CH
)
D(BS)
V. Near
Near
Medium
Far
V. Far
V. Near
Near
Medium
Far
V. Far
Normal
Normal
Normal
Bad
V. Bad
Normal
Bad
Bad
V. Bad
V. Bad
Bad
V. Bad
V. Bad
V. Bad
V. Bad
Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
V. Bad
Table 4. IF-T
HEN rule
s, where e
n
e
r
gy is mediu
m
D(CH
)
D(BS)
V. Near
Near
Medium
Far
V. Far
V. Near
Near
Medium
Far
V. Far
Good
Good
Good
Normal
Normal
Good
Good
Good
Normal
Normal
Good
Normal
Normal
Normal
Bad
Normal
Normal
Bad
Bad
V. Bad
Normal
Bad
Bad
V. Bad
V. Bad
Table 5. IF-T
HEN rule
s, where e
n
e
r
gy is high
D(CH
)
D(BS)
V. Near
Near
Medium
Far
V. Far
V. Near
Near
Medium
Far
V. Far
V. Good
V. Good
V. Good
Good
Normal
V. Good
V. Good
V. Good
Good
Normal
Good
Good
Good
Normal
Normal
Normal
Bad
Bad
V. Bad
V. Bad
Normal
Normal
Bad
V. Bad
V. Bad
Table 6. IF-T
HEN rule
s, where e
n
e
r
gy is very high
D(CH
)
D(BS)
V. Near
Near
Medium
Far
V. Far
V. Near
Near
Medium
Far
V. Far
V. Good
V. Good
V. Good
V. Good
Good
V. Good
V. Good
V. Good
Good
Good
V. Good
V. Good
Good
Good
Good
Normal
Normal
Normal
Bad
Bad
Normal
Normal
Bad
V. Bad
V. Bad
4. Performan
ce Ev
aluation
Usi
ng this n
e
twork op
eratio
n model allo
ws t
he network lifetime metric to be mea
s
ured in
data
colle
ctio
n round
s till t
he very
first
node
ru
ns
ou
t of ene
rgy.
This metri
c
i
s
kno
w
n
a
s
f
i
rst
node d
eath (FND). It has been u
s
e
d
extensively in
literatures
[4, 13, 14, and 15]. Figu
re 6
shows the flow
chart of the pr
oposed
method that i
s
a impr
oving SEP in heterogeneous
WSNs
by fuzzy app
roach.
4.1. Simulation Setup
Simulation
s a
r
e
ca
rri
ed
out
in MAT
L
AB
R20
11a
(ve
r
sion 7.1
0
) un
d
e
r
Win
d
o
w
s
10
(64
bits). The ex
perim
ents a
r
e perfo
rmed
on a PC (Thi
nk Pad E43
1
, China
)
with
an Intel W Core™
i5 Processo
rrunnin
g
at 2.
6 GHz and 4 G
B
of RAM.
For
ou
r p
r
o
p
o
se
d, 10
0
sensor no
de
s are
ra
n
doml
y
deploye
d
i
n
the
area, t
h
is
area
assume
that
(
∝
=3, P
=
0.
2, m=0.3).
Whe
r
e
∝
,
m are
co
nst
ant value
s
f
o
r
heteroge
n
e
ity
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Lifetim
e
Prolonging for
Clustere
d Heterogeneous WS
Ns by SEP-F
U
ZZY
(Yu Li)
386
percenta
ge o
f
nodes tha
n
are adva
n
ce
d, and P is
an optimal Ele
c
tion Pro
babi
lity of a node to
become
clu
s
t
e
r h
ead.Th
e t
opog
rap
h
ical
area
is
di
st
rib
u
ted in th
e di
mensi
on
100
m × 1
00 m. t
h
is
area h
a
s the
sen
s
ed tra
n
s
missio
n limit of 30 m. The perfo
rman
ce of the prop
ose
d
method
is
tested in the
s
e in the a
r
e
a
. There i
s
o
n
ly one data
sink
whi
c
h l
o
cate
d at (50
m, 50 m). The
simulatio
n
wa
s perfo
rme
d
for 200
0 roun
ds. We u
s
e a
simplified mo
del sho
w
e
d
in figure 5 for the
radio h
a
rdwa
re energy dissipation. Table
7 pr
esents th
e system
s pa
ramete
rs in d
e
tails.
Table 7. Simulation pa
ram
e
ters
Parameter Value
Topogr
aphical Area (mete
r
s)
Sink location (meters)
Number of
nodes
Limit of transmission distance (meters)
Initial energ
y
of n
ode
Packet data size
No. of MFs (in ea
ch input and outp
u
t variable)
No. of IF
-THE
N r
u
les
No. of tra
n
smission packets (roun
ds)
100m
100m
50m
50m
100
30m
0.5J
41
0
bits
5
125
2
1
0
Figure 6. Flow c
h
art of SEP - FUZZY c
l
us
tering
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 2, May 2016 : 380 –
389
387
4.2. Simulati
on Results
The n
u
mb
er of alive n
o
des a
s
a fu
nction
of ro
und
s by u
s
i
ng the
thre
e
different
approa
che
s
for this area
shown in Figure
7. It can be se
en
that the propo
sed meth
od
outperforms
FUZZY_ SEP. the network
lifetime ac
hieved by the
propos
ed
method inc
r
eas
ed
by
nearly 2
7
% than that obta
i
ned by Le
ach and al
so
i
n
cre
a
sed by n
early 23% th
an that obtai
ned
by SEP.
More
over i
n
Figure 7, it
can b
e
seen
that
the
numb
e
r of
alive n
o
des of the
propo
sed
method is always higher than both LEACH
and SEP. The different duration of time
corre
s
p
ondin
g
to the first
dead
nod
e computed
u
s
in
g the two different
app
roa
c
he
s a
r
e li
ste
d
in
Table 8. Cl
ea
rly, the time for the first no
de to
die in the propo
sed
method i
s
m
u
ch lo
nge
r th
an
the times for
the first node to die in Leach and
SEP. From T
able 8 and Figure 7 it is clear that,
the propo
se
d metho
d
o
u
tperfo
rms S
EP in term
s of bal
an
cin
g
en
ergy
co
nsum
ption
a
nd
maximizatio
n
of network lifetime.
Table 8. Nu
m
ber of ro
und
s with the first dead n
ode
Approaches
Leach
SEP
Proposed
Lifetime of the fir
s
t dead node (
R
o
unds) 186
519 91
1
Figure 8
sho
w
the ave
r
ag
e rem
a
ining
energy
of a
WSN a
s
a fu
nction
of tran
smissio
n
roun
ds fo
r the three ap
proache
s. As the rou
nd nu
mber in
crea
ses in the are
a
, the propo
sed
method performs better
than
both approaches
(LEA
CH and SEP) protocols
.
This
indicates that
,
better e
nergy balan
ce
in
a WS
N i
s
a
c
hieved
by th
e propo
se
d
method.Th
e
delay in
cu
rre
d in
transmissio
n
of data pa
ckets is
also a
key pa
ram
e
ter for
ce
rtain
appli
c
ation
s
. The comp
ari
s
on
betwe
en thre
e different a
ppro
a
che
s
is shown in
Figure 9. It can be see
n
that, the propo
sed
method has
shortest delay compared
to both
Leach and SEP. Shorter
delay indicates both
energy savin
g
and efficien
t information transmiss
io
n (espe
c
ially se
cure and imp
o
rtant one
s).
In
other words, data
pa
ckets are
r
outed th
rough diffe
rent
node
-di
s
joint
paths
with m
u
ltipath ro
uting
to avoid network
con
g
e
s
tio
n
and prolon
g
the network l
i
fetime.
Figure 1
0
sho
w
the
comp
arison
for net
work
lifetime
b
e
twee
n th
ree
different
app
roache
s.
It can be seen that, the proposed method mo
re lifetime com
pared to both Leach and SEP.
Figure 7. Nu
mber of alive
node
s a
s
a function of
roun
ds b
a
sed
on different a
ppro
a
che
s
(L
each,
SEP, and proposed).
Figure 8. Average
remai
n
in
g energy as a
function of tra
n
smi
ssi
on ro
und ba
se
d on
different approaches (Leach, SEP, and
prop
osed
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Lifetim
e
Prolonging for
Clustere
d Heterogeneous WS
Ns by SEP-F
U
ZZY
(Yu Li)
388
Figure 9. Dat
a
transmissio
n delay (si
m
u
l
ation
time for all pa
ckets) as a fu
nction of tran
smissio
n
roun
d ba
sed
on different a
ppro
a
che
s
(L
each,
SEP, and proposed)
Figure 10. Lifetime com
parison b
a
sed o
n
different approaches (Leach, SEP, and
prop
osed
)
Note th
at ab
ove sim
u
lati
ons are pe
rf
or
me
d a
s
su
ming that
all the no
de
s
are
well
maintaine
d
(i
.e. stable
with eno
ugh p
o
w
er) u
n
til the node di
es. I
n
real
wo
rld,
there m
a
y be
certai
n situati
ons that on
e even
more of the sen
s
o
r
s i
n
the crit
i
c
al
pathway become intermitt
ent
in the ability to function n
o
r
mally. Such
behavio
r may
add perfo
rm
ance noi
se (fl
u
ctuatio
ns) in
to
the WSN. As
there a
r
e too
many
param
eters to b
e
consi
dered, futu
re inve
stigat
ions a
bout su
ch
topics may be quite intere
sting an
d ch
a
llenging.
5. Conclusio
n
In ord
e
r to
im
prove th
e e
n
e
rgy effici
en
cy and a
c
hi
eve the n
e
two
r
k load
bala
n
ce
to SEP
in WS
Ns,
we
have propo
se
d a ne
w
clu
s
tering
sche
me
based o
n
fuzzy logi
c in thi
s
pa
per.
Wh
e
r
e
each no
de
d
e
termin
es its
fitness value
to be
come
cl
uster-h
ead
candid
a
te ba
sed o
n
remain
ing
energy, dista
n
ce to
cl
uste
r he
ad from
node
s a
nd di
stan
ce to
ba
se
station fro
m
clu
s
ter he
ads.
The p
e
rfo
r
ma
nce
of the
propo
sed
meth
od i
s
eval
u
a
ted u
nde
r the
sam
e
crite
r
i
a
an
d
comp
a
r
ed
with LEACH,
SEP. Simulat
i
on res
u
lts
demons
trate
t
he effec
t
iveness
of the new approac
h
with
rega
rd
s to e
nhan
cem
ent
of the lifetime of wi
re
le
ss sen
s
o
r
n
e
tworks
with ran
domly scattered
node
s.
Ackn
o
w
l
e
dg
ement
This
work
wa
s su
ppo
rted i
n
part by the
National
Nat
u
ral Sci
e
n
c
e
Found
ation o
f
China
unde
r G
r
ant
s 6147
140
8; b
y
the Natio
n
a
l
High
Te
chn
o
logy Resea
r
ch a
nd
Devel
opment P
r
og
ram
of China u
n
d
e
r Grants 2
0
1
4
AA01A701.
Referen
ces
[1]
Akka
ya K, You
n
is M. A surve
y
o
n
routi
ng pr
otocols for
w
i
r
e
less se
nsor n
e
t
w
o
r
ks.
Ad ho
c netw
o
rks
.
200
5 3(3): 32
5
-
49.
[2]
Heinz
e
lm
an W
B
, Chan
drak
as
an AP, Bal
a
kri
s
hna
n H. An
a
pplic
atio
n-sp
ec
ific
pr
otoc
ol arc
h
itecture for
w
i
rel
e
ss micro
s
ensor n
e
t
w
ork
s
.
IEEE
Transactions on Wireless Comm
unic
ations
. 20
02; 1
(
4): 660-7
0
.
[3]
Lin
d
se
y S, Ra
ghav
en
dra CS.
PEGASIS: Po
w
e
r-effici
e
n
t g
a
theri
ng i
n
sen
s
or informati
on
s
y
stems.
In
IEEE Aerospac
e conference proceedings
. 20
02; 3: 3-11
25.
[4]
Youn
is O, F
a
h
m
y
S.
HEED:
a h
y
br
id,
ener
g
y
-
e
fficie
n
t, di
stributed
clust
e
rin
g
a
ppro
a
c
h
for a
d
ho
c
sensor n
e
t
w
ork
s
.
IEEE
Transactions on Mobile Com
p
uting
2
004; 3(4): 3
66-
79.
[5]
Qing L, Z
hu Q, W
ang M. Design of a distribut
ed
ener
g
y
-efficie
n
t
clustering a
l
gorit
hm for
hetero
g
e
neo
us
w
i
r
e
less s
ens
or net
w
o
rks.
C
o
mputer co
mmunic
a
tions
. 2
0
0
6
; 29(12): 2
230
-7.
[6]
Smaragdakis
G, Matta I, Bestavros A. SEP: A st
able
election pr
otocol for
cluster
ed heterogeneous
w
i
reless sensor net
w
o
rks.
In
Secon
d
i
n
tern
ation
a
l w
o
rksh
op o
n
se
nsor
and
actor n
e
tw
ork protoc
ol
s
and a
p
p
licati
o
n
s
(SANPA 200
4).
2004; 3(
22)
.
[7]
Bala M, A
w
as
thi L. Profic
ient D-SEP prot
ocol
w
i
t
h
heterogeneit
y
for
maxi
mizing the lifetime of
w
i
reless sens
or net
w
o
rks.
Internati
ona
l Jour
nal of Intell
ig
en
t systems and
app
licati
ons.
2
012; 4(7): 1.
[8]
Aderohunmu F
A
, Deng JD.
A
n
enhanced St
able Election P
r
otocol
(SEP) f
o
r clustered heterogeneous
WSN.
Depart
m
ent of Informati
on Sc
ie
nce, Un
iversity of Otago
, Ne
w
Z
eal
an
d. 2009.
0
200
400
600
800
1000
LEACH
SEP
S
EP-FUZZY
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
2, No. 2, May 2016 : 380 –
389
389
[9]
Benkir
ane
S,
Beni
hssan
e
A,
Hasn
ao
ui ML,
Lag
hd
ir M. Di
stance-b
a
se
d
stabl
e
el
ection
protoco
l
(DB-
SEP) for heterogeneous
w
i
r
e
less sens
or net
w
ork
. Intern
ati
ona
l Journ
a
l o
f
Comp
uter Ap
plicati
ons
.
201
2; 58(1
6
).
[10]
Z
adeh LA. Soft
computin
g an
d fuzz
y
l
ogic.
IEEE software
.
199
4; 11(6): 48
.
[11]
Kulkar
ni RV,
F
ö
rster A, Ve
na
ya
gamo
o
rth
y
GK. C
o
mpu
t
ationa
l i
n
tell
ig
ence
in
w
i
re
l
e
ss sens
o
r
net
w
o
rks: a sur
v
e
y
.
Communications Surv
ey
s & Tutorials, IEEE.
2011; 13(
1): 68-96.
[12]
Runkl
e
r T
A
. Selecti
on
of ap
propri
a
te
defu
zzifica
tio
n
met
hods
usi
ng
ap
plicati
on s
pec
ific pro
perties
.
IEEE Transactions on Fu
z
z
y
System
s
. 19
97
; 5(1):72-9.
[13]
Ye M, Li C, Ch
en G, W
u
J. EECS: an e
nerg
y
e
ffici
ent cl
usterin
g
schem
e i
n
w
i
re
less s
e
n
s
or net
w
o
rks
.
InPerformanc
e
,
Comp
utin
g,
and C
o
mmun
i
c
ations C
onfe
r
ence,
20
05. IPCCC 20
05.
24th IEEE
Internatio
na
l.
2005: 53
5-5
40.
[14]
Aslam N, Phil
li
ps W
,
Robertson W
,
Sivaku
mar
S. A multi-criterion
optim
iz
atio
n techni
q
ue for ener
g
y
efficient cluster
formation in
w
i
reless se
nsor n
e
t
w
o
r
ks.
Informati
on F
u
si
on.
2011; 1
2
(3): 2
02-1
2
.
[15]
Mheme
d
R, Aslam N, Phil
lips
W
,
Comeau F
.
An
ener
g
y
effic
i
ent fuzz
y
lo
gic
cluster formati
on protoc
ol
in
w
i
reless sensor net
w
o
rks.
Proced
ia Co
mputer Scie
nce
.
201
2; 10: 255-
62.
Evaluation Warning : The document was created with Spire.PDF for Python.