Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
1
3
,
No.
1
,
Jan
uar
y
201
9
,
pp.
1
79
~
1
85
IS
S
N:
25
02
-
4752
, DO
I: 10
.11
591/
ijeecs
.
v
1
3
.i
1
.pp
1
79
-
185
179
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
The con
diti
on monitorin
g of dies
el engine
s
usin
g acousti
c
sign
al analysis
Widi
Pra
se
t
yo, Mudrik
Alay
drus
El
e
ct
ri
ca
l
Eng
in
ee
ring
Major
,
Fa
cul
t
y
of Enginee
ring,
Mer
cu
Bu
a
na
Univer
si
t
y
,
Meru
y
a, Ja
kar
ta,
I
ndonesia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
y
25
, 201
8
Re
vised
Sep
28
, 2
01
8
Accepte
d
Oct
9
, 2
01
8
Engi
ne
hav
e
becom
e
one
of
the
supportive
asset
f
or
m
an
y
activi
t
y
and
work,
the
ref
or
e
eng
ine
nee
d
a
tt
en
ti
on
for
it
s
condition
.
The
ea
si
est
wa
y
to
do
is
through
the
ac
o
ustic
sound
of
the
engi
ne
i
tsel
f
.
Mos
t
of
the
ti
m
e,
the
engi
n
e
sounds
are
checke
d
using
a
tr
adi
ti
on
al
wa
y
t
hat
m
a
y
ca
usin
g
a
deba
t
e
reg
ard
ing
it
s
co
ndit
ion.
Th
is
is
due
to
no
supportive
s
ci
en
ti
fi
c
b
asis
to
know
about
engi
n
e
co
ndit
ion
using
th
ei
r
own
ac
ousti
c
sound.
A
val
ue
from
it
s
fre
quency
p
at
t
e
rn
is
nee
ded
a
s
scie
nti
fi
c
bas
is
to
det
e
rm
ine
an
engi
n
e
condi
ti
on
using
ac
ousti
c
sound.
Method
of
env
el
ope
ext
r
ac
t
ion
b
y
h
il
b
ert
tra
nsform
,
fast
fourie
r
tra
nsform
,
and
cor
re
la
t
ion
coe
fficie
nt
are
used
to
find
tha
t
v
al
u
e.
A
se
rie
s
of
t
ests
hav
e
be
en
c
arr
i
ed
o
ut
on
th
e
va
lue
s
tha
t
hav
e
bee
n
found
an
d
the
result
ar
e
prom
ising
fo
r
te
ll
ing
engi
ne
condi
ti
on
,
but
unfortun
at
e
l
y
the
val
u
es
has
not
be
en
ab
le
to
ide
nti
f
y
t
y
p
e
of
damage
th
a
t
occ
ur
on
th
e eng
ine
.
Ke
yw
or
d
s
:
Acousti
c s
ound
Correl
at
ion
c
oe
ff
ic
ie
nt
En
velo
pe
e
xtra
ct
ion
Fast f
ourier t
ra
ns
f
or
m
Fr
e
qu
e
ncy
patte
rn
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
:
W
i
di P
rasety
o,
Ele
ct
rical
En
gi
neer
i
ng Maj
or,
Faculty
of E
ngineerin
g,
Me
rcu
B
ua
na Un
i
ver
sit
y, Me
ru
ya
,
Jak
a
rta,
I
ndonesi
a.
Em
a
il
:
widi.p
r
aset
yo@outl
oo
k.
c
om
1.
INTROD
U
CTION
En
gin
e
ha
ve
be
en
us
e
d
as
s
upportive
as
s
e
t
on
daily
li
fe.
It
su
pp
or
t
a
vast
area,
f
r
om
el
ect
rical
gen
e
rato
r
to
m
ining
eq
uipm
ent.
It
is
not
ov
er
react
if
eng
i
ne
co
ndit
ion
nee
d
to
be
m
ai
ntained
so
it
can
op
e
rate
pr
op
e
r
ly
.
Eng
i
ne
m
ain
te
na
nce
ca
n
be
de
vid
e
in
to
t
wo
ty
pe
of
m
ai
ntenance
,
the
y
are
predete
r
m
ined
pr
e
ve
ntive
m
a
intenanc
e(
pdP
M)
an
d
co
ndit
ion
ba
sed
m
ai
ntenan
c
e(CB
M)
[1
]
.
E
ngin
e
m
ai
ntenan
ce
us
in
g
pdPM
ty
pe
is
m
ai
ntenan
ce
th
at
done
pe
rio
dica
ll
y
us
ing
p
re
-
set
tim
e
interval
to
pr
e
ven
t
a
ny
m
ajo
r
dam
a
ge
of
eng
i
ne
com
ponen
t
f
ro
m
occu
r
.
O
n
the
othe
r
ha
nd,
en
gine
m
ai
ntenan
ce
us
in
g
CB
M
is
fo
c
us
on
m
on
it
or
in
g
syst
e
m
of
en
gin
e
co
ndit
ion
w
hich
regulary
t
ran
sm
it
a
6
n
inf
or
m
at
ion
that
at
so
m
e
po
int
th
os
e
in
form
at
io
n
ar
e
colle
ct
ed
on
a
par
am
et
er
that
can
sta
te
the
e
ng
i
ne
c
onditi
on.
CB
M
ty
pe
ha
ve
3
keys
el
e
m
ent
to
determ
ine
the
m
ai
ntenan
ce
of
the
m
on
it
ored
en
gi
ne.
T
hose
keys
are
data
colle
ct
io
n,
data
proce
ssing,
an
d
de
ci
sion
m
aking
[
1
-
2].
Ther
e
are
som
e
adv
a
ntage
s
that
can
be
ob
ta
in
by
doin
g
e
ng
i
ne
co
nd
it
io
n
m
on
it
or
ing.
Th
os
e
ad
van
ta
ges
are
dec
reas
ing
m
ai
ntenance
cost
of
an
e
ng
i
ne
as
far
as
50
%
-
80%,
de
creasin
g
the
a
m
ou
nt
of
e
quipm
ent
dam
ages
as
far
as
50%
-
80
%,
dec
reasin
g
ov
e
rtim
e
costs
of
hum
an
res
ources
as
far
a
s
20%
-
50%,
inc
reasin
g
li
fetim
e
of
a
n
en
gin
e
a
s
fa
r
as
50%
-
60
%,
an
d
inc
reas
ing
the
t
otal
pro
duct
ivit
y
as
far
as
20%
-
30%
[1], [
3].
The
f
ast
est
an
d
easi
est
way
to
determ
ine
engi
ne
co
ndit
ion
i
s
thr
ough
it
s
a
coust
ic
sound.
It
is
can
be
done
beca
us
e
eng
i
ne
aco
us
ti
c
so
und
ca
n
re
la
y
an
info
rm
at
ion
re
gardin
g
it
s
con
diti
on
[
4
-
7].
As
f
or
e
xam
ple
that
us
in
g
sou
nd
as
s
ource
of
inf
or
m
at
ion
are
unde
rw
at
e
r
ta
rg
et
recog
niti
on
[
8
-
9
]
.
Ac
ou
sti
c
sou
nd
t
hat
are
pro
du
ce
by
eng
ine
can
be
cat
egorize
as
“he
al
thy”
if
there
are
n
o
k
noc
king
sound,
ratt
li
ng
sound,
a
nd
huntin
g
so
un
d
[9
]
. T
he
refor
e
engine
c
onditi
on ca
n b
e know
n by usi
ng aco
us
ti
c s
ound
of the
engi
ne.
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,
Vol
.
1
3
, N
o.
1
,
Ja
nu
a
ry
201
9
:
1
7
9
–
1
8
5
180
Most
of
the
ti
m
e,
eng
ine
co
nd
it
io
n
m
on
it
or
in
g
thr
ough
acoust
ic
so
und
are
done
in
tra
diti
on
al
way
,
wh
ic
h
ca
n
le
ad
to
a
de
bate
re
gardin
g
t
he
e
ngine
c
onditi
on
[2
]
.
It
is
occur
because
la
ck
of
sci
entic
fou
ndat
io
n
that
suppo
rt
th
e
recog
niti
on
of
en
gin
e
co
ndit
ion
us
in
g
it
s
own
aco
us
ti
c
s
ound.
Be
side
t
ha
t,
it
rely
on
pe
rson’s
long
tim
e
exp
eriences
in
e
ngine
m
ai
ntenan
ce
div
isi
on
a
r
ea
and
that
pe
rson’s
hea
rin
g
abili
ty
.
Ther
efore
the
resu
lt
of en
gine
m
on
it
or
in
g w
il
l be sub
j
ect
iv
e
[
2].
Thro
ugh
this
r
esearch
e
ngine
so
un
d
will
be
visu
al
iz
e
as
fr
e
qu
e
ncy
patte
r
n
and
c
om
par
e
tho
s
e
patte
r
n
betwee
n
healt
hy
en
gin
e
a
nd
dam
aged
e
ng
i
ne.
Ne
xt,
t
ho
s
e
f
re
qu
e
nc
y
patte
rn
will
be
us
e
d
t
o
fin
d
a
char
act
e
risti
c
value
that
can
disti
ng
uis
h
he
al
thy
eng
ine
f
r
om
da
m
aged
eng
i
ne.
W
it
h
th
is
fr
eq
uen
cy
pa
tt
ern
dan cha
racteri
s
ti
c v
al
ue
c
ondi
ti
on
of en
gin
e
can
be
m
ai
ntain
ed
.
Si
m
il
ar
resear
ch
hav
e
bee
n
cond
ucted
se
ve
r
al
tim
e
us
ing
va
rio
us
m
eth
od.
Am
ong
them
wav
el
et
trans
form
[5
]
,
[
8
-
1
1
]
is
m
o
re
popula
r
the
n
the
oth
e
r.
Othe
r
m
et
ho
d
su
ch
as
li
nier
pr
e
dicti
on
a
naly
sis
m
et
ho
d
[
4],
blind
s
ource
se
pa
rati
on
(B
SS)
m
et
hod
[
3],
ho
m
om
or
phic
analy
sis
m
e
tho
d
[9]
and
Hilbert
-
Hu
a
ng
Transf
or
m
(HHT)
[
12]
ha
ve
al
so
bee
n
use
d
to
c
onduct
resea
rch
re
ga
rd
i
ng
aco
us
ti
c
sou
nd
of
t
he
en
gin
e
conditi
on.
Hon
gj
ia
ng
He
,
Chun
xia
W
a
ng,
a
nd
Yaz
hou
cond
uct
a
rese
arch
t
o
detect
a
com
po
ne
nt
dam
age
us
i
ng
wav
el
et
tra
nsf
or
m
by
isolat
ing
a
s
ound
s
ource
of
dam
aged
com
pone
nt.
T
he
resu
lt
are
s
uccess
ful
with
con
cl
us
io
n
t
hat w
a
velet
tran
s
f
or
m
h
ave
m
or
e resista
nce
to
e
nv
i
ronm
ent n
oi
se [5]
.
R.M
.
Vile
la
,
J
.C.
Me
trol
ho,
and
J.C.
Ca
rdoso
co
nduct
r
esearch
to
e
xtract
a
sin
gle
s
ign
al
s
ound
so
urce
f
r
om
a
m
ixed
signa
l
so
un
d
sou
rce
usi
ng
BS
S
m
e
tho
d
T
DSEP
al
gorithm
.
Althou
gh
t
his
resea
rc
h
are
done
in
a
co
ntr
olled
en
vir
on
m
ent
la
borato
ry
us
in
g
a
si
m
ulati
on
of
dam
age
d
en
gin
e
c
om
po
ne
nt,
the
re
sul
t
are
prom
isi
ng
[3].
Li Jun, Dong Xin
gw
e
n,
Son
g
Ya
j
i, a
nd
S
ong N
uan
c
ondu
ct
r
esearc
h
to
e
xtract an
e
ng
i
ne
so
un
d
f
ro
m
an
en
vironm
ent
no
ise
with
hom
o
m
or
phic
analy
sis
m
et
ho
d
co
ntinu
e
d
w
it
h
cepstru
m
lin
es
to
determ
i
ne
the
eng
i
ne
co
ndit
ion.
Re
su
lt
of
the
resea
rc
h
sa
id
that
with
m
or
e
ce
ps
tr
um
l
ines
the
acc
ura
cy
to
recog
nize
the
eng
i
ne
c
onditi
on
wi
ll
inc
reas
e as
well
[9
]
.
Sadee
p
Ku
m
ar
Yad
a
v,
Ka
ni
sh
ka
Ty
agi,
Brijes
hkum
ar
Sh
a
h,
a
n
d
Pr
e
m
Ku
m
ar
Kal
ra
c
onduct
researc
h
ot
ide
ntify
en
gin
e
c
onditi
on
usi
ng
e
nv
el
op
e
detect
ion
a
n
d
FFT.
Correl
at
ion
c
oe
ff
ic
ie
nt
are
use
d
f
or
m
aking
a
deci
sion
w
hethe
r
the
en
gin
e
i
n
healt
hy
conditi
on
or
dam
aged
co
nd
it
io
n.
Using
those
m
et
ho
d
su
ccess
fu
ll
y re
cognize a
n
e
ng
ine d
am
age typ
e up to
80%
succ
ess r
at
e
[2
]
.
Qian
g
W
a
ng
and
X
uem
in
Tia
n
co
nduct
researc
h
of
s
oft
-
se
ns
in
g
al
go
rithm
us
ing
H
il
ber
t
-
H
ua
ng
Transf
or
m
(HHT)
a
nd
Wav
e
le
t
Su
pp
ort
Ve
ct
or
Ma
c
hin
e
(
WSVM
).
The
app
li
cat
io
n
of
HH
T
-
WSVM
are
use
d
as f
eat
ure extra
ct
ion
as p
a
rts of soft
-
se
ns
in
g
a
lgorit
hm
. Th
e r
esearch res
ult suggeste
d
that u
sing
HHT
-
WSVM
are
m
or
e
accur
at
e
com
par
ed
t
o
ot
her
t
wo
m
et
hods
,
KP
C
A
-
LSS
VM
a
nd
HH
T
-
S
VM.
T
his
m
eans
that
HHT
-
WSVM
h
a
s a
be
tt
er p
re
dicti
on
and
gen
e
rali
zat
ion
[12].
Zh
ongjie
W
a
ng,
Ji
ngna
n
Z
ha
ng,
a
nd
Ya
ngch
un
Lia
ng
cond
uct
a
re
s
earch
t
o
te
st
no
ise
s
an
d
vibrat
ion
s
of
a
m
oto
r.
The
F
FT
analy
sis
m
et
hod
is
us
ed
t
o
acq
uire
the
no
ise
sig
nal,
r
ecordin
g
an
d
s
pectral
analy
sis.
T
his
researc
h
c
oncl
ud
e
that
m
oto
r
noise
a
nd
vi
brat
ion
from
this
resea
rch
can
be
us
e
d
to
rec
ognize
the con
diti
on of a
no
t
her
m
otor [7]
.
Using
ab
ove
r
esearch
jour
na
l
[2
]
as
an
ins
pirati
on,
this
r
esearch
a
re
usi
ng
e
nvel
ope
de
te
ct
ion
a
nd
FFT
to
get
t
he
aco
us
ti
c
sou
nd
f
reque
ncy
patte
rn.
T
he
de
ci
sion
m
akin
g
are
usi
ng
c
om
par
ison
bet
ween
fr
e
qu
e
ncy
patte
rn of
healt
hy
eng
i
ne
ac
ousti
c sou
nd and
da
m
aged
e
ng
i
ne a
coust
ic
sound
.
2.
RESEA
R
CH MET
HO
D
As
ge
neral
picture,
the
rese
arch
m
et
ho
d
t
hat
are
us
e
d
c
on
sist
of
se
ve
ral
ste
p
from
sam
pling
to
decisi
on
m
aki
ng.
The
first
st
ep
are
sam
pling
the
ac
ou
sti
c
so
un
d
of
a
n
en
gin
e.
The
n
c
onti
nu
e
to
proces
s
the
sam
pling
us
in
g
MA
TLAB
s
of
t
war
e
t
o
gain
the
i
ns
ta
nta
neous
am
plit
ud
e
a
nd
t
he
n
usi
ng
FFT
t
o
ge
t
the
fr
e
qu
e
ncy
patte
rn.
The
decis
ion
m
aking
process
are
ba
s
ed
on
correla
ti
on
coe
ff
ic
ie
nt
on
com
par
is
on
of
fr
e
qu
e
ncy
patt
ern
of
healt
hy
eng
i
ne
aco
us
t
ic
so
und
a
nd
dam
aged
en
gin
e
aco
us
ti
c
sound.
Fi
gure
1
sh
ow
s
il
lustrati
on
of a
bove st
eps.
SO
U
N
D
R
E
C
O
R
D
D
O
W
N
S
A
M
P
L
I
N
G
E
N
V
E
L
O
P
E
D
E
T
E
C
T
I
O
N
by
H
I
L
B
E
R
T
T
R
A
N
SF
O
R
M
FFT
C
O
R
R
E
L
A
T
I
O
N
C
O
E
F
F
I
C
I
E
N
T
Figure
1. Re
se
a
rch ste
p diag
r
a
m
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
Th
e c
onditi
on
monitori
ng o
f
diesel
engine
s
us
in
g ac
ou
sti
c
signal
analysis
(
Wi
di Praset
yo
)
181
Sam
pling
is
done
by
c
ollec
ti
ng
e
ngine
ac
ou
sti
c
s
ound
t
hat
ope
rate
at
sp
ee
d
1.5
00
r
pm
,
no
l
oad
,
and
us
in
g
only
s
m
artphon
e
m
ic
ropho
ne.
E
ac
h
sam
ple
of
en
gin
e
ac
ou
s
ti
c
s
ound
are
rec
orded
with
5
sec
onds
durati
on
an
d
t
aken
at
20
c
m
fr
om
eng
ine
body.
Since
sm
artph
one
m
ic
ro
phone
m
os
tl
y
hav
e
22.
000
Hz
fr
e
qu
e
ncy
rang
e,
the
sam
pling
fr
e
qu
e
ncy
is
set
at
44.
100
Hz
a
nd
t
he
rec
ordin
g
is
save
d
in
WAVE
(
wa
v)
file
.
The
source
of
sam
ple
that
wa
s
colle
ct
ed
are
from
var
io
us
e
ng
i
ne
m
aker
a
nd
m
od
el
.
T
his
is
dee
m
ed
nec
essary
in
orde
r
to
m
a
ke
the
e
ng
i
ne
m
anu
fact
ur
e
r
a
nd
m
od
el
not
lim
i
te
d
j
ust
one
kind.
Ta
b
le
1
s
how
s
t
he
am
ou
nt
of
ta
ken
sam
ple an
d
T
a
bl
e
2
s
how
s
th
e s
ource
s
of en
gin
e
sou
nd sam
ple
s
.
Table
1.
Sam
pl
e
Q
uan
ti
ty
So
u
n
d
T
y
p
e
Qu
an
tity
Health
y
So
u
n
d
36
Kn
o
ck
in
g
Sound
23
Hu
n
tin
g
Sou
n
d
26
Table
2.
Sour
c
e of
E
ng
i
ne
S
ound
Sam
ple
Manu
f
actu
rer
Mod
el
Ru
n
n
in
g
Hou
r
Ap
p
licatio
n
Cap
acity
CATERP
IL
L
AR
D3
4
1
2
5
5
.11
3
GENS
ET
8
1
0
k
VA
CATERP
IL
L
AR
D3
5
1
2
3
9
.66
2
COMPRES
SOR
7
5
0
k
VA
CATERP
IL
L
AR
D3
3
0
6
4
2
.44
1
GENS
ET
3
5
0
kVA
PERKIN
S
2
0
0
6
TT
AG
1
2
.77
3
GENS
ET
4
5
0
k
VA
NISSA
N
RD1
0
1
4
.26
3
GENS
ET
2
5
0
kVA
VOLV
O
TAD
1
0
3
0
1
5
.06
0
GENS
ET
2
5
0
kVA
Nex
t,
eac
h
sa
m
ple
wil
l
go
thr
ough
series
of
process
un
ti
l
the
fr
eq
ue
ncy
patte
rn
is
f
orm
ed.
Since
the
or
i
gin
al
sam
pl
e
hav
e 5
seco
nds
durati
on
a
nd
44.
100Hz
sa
m
pl
ing
f
reque
ncy,
it
gen
e
rate
approxim
at
ely
22
.
000
data
po
i
nt
in
MATLAB
.
To
ease
the
com
pu
ta
ti
on
pro
ce
ss
and
s
horten
the
processin
g
tim
e,
on
ly
8.800
dat
a
po
i
nts
or
e
qual
to 5 ti
m
es engine rota
ti
on w
il
l go th
rou
gh n
e
xt pr
ocess.
The
8
.
800
data
points
a
re
pro
cessed
with
hil
ber
t
t
ran
s
f
or
m
to
get
the
a
nal
yt
ic
sign
al
,
a
nd
by
do
i
ng
that
the
real
value
d
sig
nal
is
conver
te
d
into
c
om
plex
sign
al
.
T
he
an
al
yt
ic
sign
al
or
com
plex
sig
nal
i
s
necessa
ry
to g
e
t
the
sig
nal’s
e
nv
el
op
e
detect
ion.
T
he
purpos
e
of
this
e
nv
el
op
e
detect
io
n
i
s
to
sig
nify
a
ny
peak
that
exist
in
th
e
com
plex
sign
al
.
T
he
com
plex
sig
nal
co
ns
i
st
of
2
pa
r
ts,
r
eal
-
value
si
gn
a
l
par
t
an
d
im
a
gin
a
ry
-
value
si
gn
al
p
a
rt.
The
real
pa
r
t
are
the
sig
nal it
sel
f
that
do
es
n’
t go
t
hroug
h
any
proces
s.
A
no
t
her
case
wit
h
the
i
m
aginar
y
par
t
,
this
par
t
is
an
ou
tp
ut
of
hilb
ert
trans
form
ation
.
T
he
hilber
t
transfor
m
at
ion
eq
uatio
n
is
wr
it
te
n
as foll
ows [1
2
-
15
]
,
[
(
)
]
=
̃
(
)
=
1
∫
(
)
−
∞
−
∞
(1)
wh
e
re
is
tim
e,
(
)
is
ti
m
e
do
m
ai
n
si
gn
al
,
a
nd
̃
(
)
is
the
re
su
lt
of
hilbert
t
rans
form
as
im
agi
nar
y
-
valu
e
sign
al
par
t.
With
real
pa
rt
a
nd
im
aginar
y
par
t
al
rea
dy
know
n,
t
hen
t
he
com
plex
sig
n
al
can
be
f
orm
us
ing
fo
ll
owin
g
E
qu
at
ion
[12
-
15]
.
(
)
=
(
)
+
̃
(
)
=
(
)
(
)
(2)
The
act
ual
f
orm
of
env
el
ope
detect
ion
that bein
g
us
e
in
t
hi
s
researc
h
is
an
abs
olu
te
va
lu
e
of
com
plex
sign
al
it
sel
f.
T
he
e
nv
el
op
e
de
te
ct
ion
or
known
as
in
sta
nta
neous
am
plit
u
de
hav
e
r
ole
to
sig
nify
any
a
m
pl
it
ud
e
relat
ed
sig
nal c
har
act
erist
ic
. I
n
c
om
plex
sig
na
l abs
olu
te
value
e
qu
at
io
n
is
wr
it
te
n
as
foll
ows
[12
-
15]
.
(
)
=
|
(
)
|
=
√
2
(
)
+
̃
2
(
)
(3)
Nex
t,
t
he
sig
na
l’s
en
velo
pe
detect
ion
wil
go
t
hro
ugh
F
F
T
process
t
o
ge
t
into
f
reque
nc
y
do
m
ai
n.
The
res
ult
of
this
FFT
proc
ess
al
ways
s
how
2
side
of
sign
al
,
that
is
po
sit
ive
val
ue
an
d
ne
gative
valu
e.
In
this
resea
rc
h,
only
posit
iv
e
val
ue
will
be
us
e
d
as
c
omparis
on
[
16]
.
The
FFT
res
ul
t
will
be
sh
own
on
log
a
rithm
fo
rm
to
sh
ow
a
ny
detai
l
that
exist
in
the
fr
e
quency
patte
r
n.
F
igure
2
giv
es
t
he
en
velo
pe
de
te
ct
ion
and FF
T r
e
su
lt
.
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,
Vol
.
1
3
, N
o.
1
,
Ja
nu
a
ry
201
9
:
1
7
9
–
1
8
5
182
Figure
2. En
ve
lop
e
detect
io
n and FF
T
The
com
par
iso
n
process
will
be
co
nducted
with
correla
ti
on
coeffic
ie
nt
m
et
hod.
the
ai
m
is
to
get
the
value
of
c
om
par
ed
si
gn
al
in
fr
e
qu
e
ncy
do
m
ai
n
and
by
us
in
g
that
val
ue
en
gin
e
c
onditi
on
ca
n
be
detect
ed
betwee
n healt
hy
an
d
dam
aged
.
T
his r
esea
rc
h i
s u
sin
g Pe
ars
on c
orrelat
ion c
oeffici
ent
,
as foll
ows [2]
,
,
=
(
(
−
)
(
−
)
)
(4)
Wh
e
re
and
is
two
ra
ndom
var
ia
bles,
and
is
the
ex
pected
values
,
and
is
the
sta
nd
a
rd
dev
ia
ti
ons,
and
is t
he
expect
e
d value
operat
or.
A
s
et
of
healt
hy
eng
i
ne
s
ound
is
pre
par
e
d
as
a
be
nc
hm
ark
that
eac
h
on
e
w
il
l
be
com
par
e
d
with
a
te
st
eng
i
ne
s
ound.
The
n,
the
outp
ut
value
of
ea
ch
com
par
iso
n
will
be
ave
ra
ged
a
nd
as
a
r
esult
,
a
sin
gle
ou
t
pu
t
value
wil
be
ge
ner
at
e
d.
This
value
is
us
ed
to
detect
i
ng
eng
i
ne
co
nd
it
io
n.
I
f
there
are
m
ulti
ple
te
s
t
engi
ne
so
un
d,
t
he
c
orr
el
at
ion
coe
ff
ic
i
ent
will
f
or
m
a
m
a
trix
that
ea
ch
c
olu
m
n
ha
ve
tho
s
e
av
era
ge
value
of
fr
e
quency
patte
rn
c
om
par
ison
[17]
.
Ev
e
n
if
the
am
ou
nt
of
healt
hy
e
ngine
s
ound
sa
m
ple
and
te
st
eng
i
ne
s
ou
nd
i
ncr
ease
,
the
bacis
m
eth
od
of
correla
ti
on
coe
ff
ic
ie
nt
in
this
research
are
sti
ll
t
he
sam
e.
Figu
re
3
il
lustrate
s
the
correla
ti
on co
e
ff
ic
ie
nt m
et
ho
d an
d form
ed
m
at
rix.
1
2
3
4
5
1
2
3
4
5
S
ou
nd
o
f
t
e
s
t
e
ng
i
ne
C
ol
l
e
c
t
i
on
of
he
a
l
t
hy
e
ngi
ne
s
ound
ρ
(
1
,
1
)
ρ
(
2
,
1
)
ρ
(
3
,
1
)
ρ
(
4
,
1
)
ρ
(
5
,
1
)
ρ
(
1
,
2
)
ρ
(
2
,
2
)
ρ
(
3
,
2
)
ρ
(
4
,
2
)
ρ
(
5
,
2
)
ρ
(
1
,
3
)
ρ
(
2
,
3
)
ρ
(
3
,
3
)
ρ
(
4
,
3
)
ρ
(
5
,
3
)
ρ
(
1
,
4
)
ρ
(
2
,
4
)
ρ
(
3
,
4
)
ρ
(
4
,
4
)
ρ
(
5
,
5
)
ρ
(
1
,
5
)
ρ
(
2
,
5
)
ρ
(
3
,
5
)
ρ
(
4
,
5
)
ρ
(
5
,
5
)
X
2
X
3
X
4
X
5
M
e
a
n
o
f
c
o
lu
m
n
X
1
Figure
3. Ma
trix
of
co
rr
el
at
io
n
c
oeffici
ents
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
Th
e c
onditi
on
monitori
ng o
f
diesel
engine
s
us
in
g ac
ou
sti
c
signal
analysis
(
Wi
di Praset
yo
)
183
3.
RESU
LT
S
A
ND AN
ALYSIS
3.1.
Fre
quen
cy
P
attern
an
d Char
acter
istic
Value
W
it
h
the
m
e
thod
m
ention
ed
a
bove,
w
e
can
get
a
fr
e
quency
pa
tt
ern
of
hea
lt
hy
eng
i
ne,
knoc
king
e
ng
i
ne,
a
nd
huntin
g
e
ng
i
ne.
T
hos
e
3
fr
e
quency
patte
rn
ha
ve
de
ff
e
ren
ces
bet
ween
them
.
For
healt
hy
eng
i
ne
the
high
powe
r
exist
on
0
Hz
–
5.0
00
Hz,
an
d
the
n
after
5.0
00
Hz
the
powe
r
are
relat
ively
l
ow
a
nd
ste
ady.
H
un
ti
ng
en
gin
e
the
hi
gh
powe
r
exist
on
0
Hz
–
5.0
00
Hz,
the
n
be
tween
5.0
00
H
z
and
15.
000
Hz
th
e
powe
r
are
low
with
so
m
e
po
we
r
increase
between
the
m
,
the
ov
er
15.
000
Hz
the
powe
r
increas
e
again
.
Knoc
king
e
ng
i
ne
the
high
power
e
xist
on
0
Hz
–
15.
000
Hz
an
d
relat
iv
el
y
ste
ady,
the
n
afte
r
15.
000
Hz
the
powe
r
dr
op
a
nd
ste
ady
unti
l
22.00
0
Hz.
T
hose
3
patte
rn
ha
ve
hi
gh
powe
r
bet
ween
0
H
z
–
5.000
Hz,
it
is
due
to
the
sou
nd
f
r
equ
e
ncy
of
en
gi
ne
sound
it
sel
f.
It
is
sai
d
that
the
eng
i
ne
sou
nd
f
re
quency
that
pe
netrate
eng
i
ne
structu
ral
body
are
a
ppr
ox
im
at
el
y 3.000 H
z
[
18
]
.
Fig
ur
e
4 a
nd F
ig
ure
5
s
how
th
os
e
3 fr
e
qu
e
ncy
patte
rn.
Figure
4. Fr
e
quency
patte
r
n of healt
hy e
ngine
(a)
(b)
Figure
5. (a
)
F
r
equ
e
ncy
patte
r
n of h
unti
ng engine,
(b) F
requ
ency patt
er
n o
f
kno
c
king e
ngine
Fr
om
fr
e
qu
e
nc
y
patte
rn
on
F
igure
4,
a
value
is
ge
nerat
ed
us
i
ng
pat
te
rn
c
om
par
ison
base
d
on
c
orrelat
ion
c
oe
ff
ic
ie
nt
m
et
h
od.
The
sam
ple
of
e
ng
i
ne
s
ound
are
bein
g
te
ste
d
a
co
uple
of
ti
m
e
to
get
a
char
act
e
risti
c
value
of
he
al
th
y
eng
ine
,
knoc
king
en
gin
e
,
and
huntin
g
en
gin
e.
T
hose
ch
aracte
risti
c
value
ar
e
0.75
f
or
healt
hy
eng
i
ne,
0.6
–
0.7
for
hunti
ng
en
gine
,
a
nd
0.4
–
0.6
5
f
or
knoc
king
en
gi
ne.
Th
os
e
val
ue
are
us
e
d
as
f
ollows:
a)
If
the
com
par
is
on
a
ver
a
ge
val
ue
ab
ov
e
0.75,
then
the
en
gine
in
healt
hy
con
diti
on.
if
the
com
par
ison
aver
a
ge value
belo
w 0.
75, the
n
the
engin
e i
n dam
aged
con
diti
on
.
b)
If
t
he
c
om
par
ison
ave
ra
ge
va
lue
bet
ween
0.6
-
0.7
,
the
n
the
e
ng
i
ne
is
in
dam
aged
c
onditi
on
a
nd
cause
d by hu
nting
.
c)
If
t
he
c
om
par
ison
a
ver
a
ge
va
lue
betwee
n
0.4
-
0.65,
t
he
n
t
he
e
ng
i
ne
is
in
dam
aged
c
onditi
on
a
nd
caus
e
d by kn
oc
king.
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,
Vol
.
1
3
, N
o.
1
,
Ja
nu
a
ry
201
9
:
1
7
9
–
1
8
5
184
3.2.
Res
earc
h
R
esul
t
Test
hav
e
bee
n
co
nducte
d
s
ever
al
ti
m
e
us
ing
the
c
ha
rac
te
risti
c
value,
as
resu
lt
f
ro
m
85
ac
ousti
c
so
un
d
si
gn
al
m
os
t
of
the
ti
m
e,
eng
i
ne
c
onditi
on
was
s
uccess
fu
ll
y
de
te
ct
ed
betwee
n
healt
hy
en
gi
ne
a
nd
dam
aged
en
gi
ne.
Unf
or
tu
nat
el
y,
fo
r
detect
ion
of
dam
age
ty
pe
the
res
ult
are
not
sat
isf
act
or
y.
T
he
re
searc
h
resu
lt
ar
e
f
rom
36
healt
hy
en
gin
e
s
ound
it
was
su
c
essfu
ll
y
detec
t
35
–
36
he
al
thy
eng
i
ne
so
un
d,
from
23
kn
ocki
ng
e
ng
i
ne
s
ound
it
w
as
s
ucc
essfu
ll
y
detect
8
–
12
knoc
kin
g
e
ngine
s
ou
nd,
f
ro
m
26
huntin
g
eng
i
ne
sound
it
was
su
ccess
fu
ll
y
detect
13
–
17
huntin
g
en
gin
e
sou
nd.
Table
3
giv
es
the
res
ult
of
the
researc
h.
Table
3.
Res
ult o
f
E
ng
i
ne
C
onditi
on
Detect
io
n
Nu
m
b
e
r
o
f
T
est
Detected
Health
y
Bro
k
en
Hu
n
tin
g
Kn
o
ck
in
g
Un
d
ef
in
ed
1
36
49
16
10
23
2
36
49
16
9
24
3
36
49
13
11
25
4
36
49
17
10
22
5
36
49
17
12
20
6
36
49
14
11
24
7
35
50
15
11
24
8
36
49
15
10
24
9
35
50
14
8
28
10
36
49
13
10
26
4.
CONCL
US
I
O
N
The
co
ncl
us
io
n
f
ro
m
this
research
with
pr
opose
d
m
et
ho
d
are
that
the
ge
ner
at
e
d
f
reque
ncy
patte
r
n
s
betwee
n
thr
ee
eng
i
ne
co
ndit
ion
s
s
howi
ng
disti
nctive
dif
f
eren
ces
.
T
ho
se
fr
e
qu
e
ncy
pat
te
rn
s
are
proc
essed
us
in
g
m
at
he
m
at
ic
al
m
et
ho
d
to
gen
e
rate
char
act
e
risti
c
va
lue
s
wh
ic
h
cou
l
d
re
veal
the
en
gin
e
c
onditi
on
betwee
n
healt
hy
e
ng
i
ne
s
or
dam
aged
e
ng
ine.
U
nfor
t
un
a
te
ly
,
tho
se
ch
aracte
risti
c
val
ue
s
sti
ll
can
no
t
give
sat
isfact
or
y
res
ult t
o detec
t t
he
ty
pe
of
dam
a
ge
e
xp
e
rience
d by a
n
e
ng
i
ne.
W
it
h
m
or
e
so
und
sam
ples
fro
m
diff
ere
nt
en
gin
e
m
anu
fact
ur
e
rs
an
d
m
odel
s,
this
m
et
ho
d
can
be
us
e
d
m
or
e
un
ive
rsal
to
detect
var
i
ou
s
e
ngine
s
ou
nd.
A
ny
im
pr
ov
em
ent
can
be
m
ade
to
m
ake
this
researc
h
m
or
e
us
ef
ul.
For
e
xa
m
ple,
by
add
i
ng
real
ti
m
e
m
on
it
or
i
ng
featu
r
e
on
sm
artpho
ne
platfo
rm
that
can
be
us
e
t
o
detec
t
an
e
ng
i
ne
c
ondi
ti
on
on t
he
s
pot.
REFERE
NCE
S
[1]
Patri
cia
Henri
qu
ez
,
Jesus
B.
Alo
nso,
Miguel
A.
Ferre
r,
&
C
arl
os
M.
Tra
vi
eso.
(2
014).
Review
of
Autom
at
ic
Faul
t
Diagnosis
S
y
ste
m
s
U
sing
Audio
and
Vibra
t
ion
Signal
s.
IE
EE
Tra
nsac
ti
on
on
S
y
stems
,
Man,
and
C
y
ber
ne
ti
cs
:
S
y
stems
,
44
(4
),
642
~ 652.
[2]
Sande
ep
Kum
ar
Yada
v,
Kanishka
T
y
agi,
Brijeshkum
ar
Shah,
&
Prem
Ku
m
ar
Kalr
a.
(2011).
Audio
Signat
ure
-
Based
Condi
ti
o
n
Monitori
ng
o
f
Inte
rn
al
Com
bustion
Engi
n
e
Us
ing
FF
T
a
nd
Corre
l
at
ion
Approac
h.
IEEE
Tra
nsac
ti
on
o
n
I
nstrum
ent
at
ion
a
nd
Mea
surem
ent,
60
(4), 1217
~
1226.
[3]
R.
M.
Vile
l
a,
J.
C.
Metrol
ho,
&
J.
C.
Cardoso.
(
2004).
Ma
chi
n
e
and
Industria
l
M
onit
orizat
ion
S
y
stem
by
Ana
l
y
s
i
s
of
Acoustic
Sign
at
ure
s.
Pro
ce
ed
i
ngs
of
the
12th
I
EE
E
Med
it
err
an
ea
n
E
le
c
trotec
hn
ic
a
l
Confer
enc
e
,
(pp.
277
~
279).
Portugal
:
Escola
Superior
d
e Tec
nologi
a
,
Insti
tuto Polit
e
cnico
d
e Caste
lo
Branc
o
.
[4]
Shuais
hi
Li
u
,
M
in
Yang,
Kep
ing
Li
u,
&
Ch
en
C
hen.
(2010)
.
Res
ea
rch
on
Feat
u
re
Ext
ra
ct
ion
of
E
ngine
Abnorm
al
Sound
Signal
B
ase
d
on
Li
ne
ar
Predic
ti
on
Anal
y
sis.
IEEE
In
te
r
nat
ion
al
Conf
er
enc
e
on
Com
put
er,
Me
chatroni
c
s,
Control
and
E
lec
troni
c
Engi
n
ee
r
i
ng,
(pp
.
76
~ 79
)
.
Chang
chun, Ch
ina
:
Changc
hun
Univer
sit
y
of
T
e
chnol
og
y
.
[5]
Hongjia
ng
He
,
Chunxia
W
ang,
&
Yaz
hou
W
u.
(2010).
Appl
ic
a
ti
on
of
W
av
el
e
t
Tr
ansform
in
Sound
Source
Dete
c
ti
on
and
F
aul
t
Diagnost
ic.
IEE
E
Int
ern
a
ti
on
al
Confer
enc
e
o
n
Com
pute
r,
Mec
hat
ron
ic
s,
Cont
rol
and
Elec
tron
i
c
Engi
ne
eri
ng,
(pp
.
298
~ 301). Ch
angc
hun,
Chin
a:
Hebe
i
Univ
ersity
of
Eng
ineeri
ng
.
[6]
Meka
la
N
,
Muniraj
C
,
Ramesh
Bal
aji
S.M.
(20
15).
Vibra
ti
on
a
nd
Noise
Anal
ysis
of
4
Φ
Sw
it
che
d
Rel
u
ct
a
n
c
e
Motor
Drive
.
T
E
LKOM
NIK
A Indone
sian
Journa
l
of
E
lectr
i
ca
l
En
gi
nee
r
ing, 14
(3)
,
410
~ 419.
[7]
Zhongj
ie
W
ang
,
Jingnan
Zh
a
ng,
Yongchun
Li
ang
.
(2013)
.
Motor
Noise
and
Vibra
t
ion
Te
st
Rese
arch.
TE
LKOM
NIK
A
Indone
sian
Jour
nal
of
E
le
c
tri
c
al
Engi
ne
eri
ng
,
11
(1),
87
~ 94.
[8]
Selva
Ba
la
n
,
Art
i
Khapa
rd
e,
Van
it
a
Ta
nk
,
T
ej
ash
ri
Rade,
&
Kirti
Ta
ka
lka
r
.
(2014)
.
Under
W
ater
Noise
Redu
ction
Us
ing
W
ave
le
t
a
nd
Savit
zk
y
-
Gol
a
y
.
Second
Int
er
nat
ion
al
Confer
e
nce
on
Com
puta
ti
onal
Sci
ence
a
nd
Engi
nee
r
ing,
(pp.
243
~ 250).
Maha
rashtr
a
,
In
dia
:
Maha
r
ashtra Insti
tute
of
T
echnolog
y
.
[9]
Li
Jun,
Dong
Xi
ngwen,
Song
Ya
ji
,
&
Song
Nuan
.
(2013).
Stud
y
o
n
Engi
n
e
Fault
Diagnosis
Based
on
the
Abnorm
a
l
Sound
Anal
y
sis
in
Tra
nsform
at
i
on
Dom
ai
n.
Fourth
IEEE
Inte
r
nat
ion
al
Confer
enc
e
on
Dig
it
a
l
Manufa
ct
ur
ing
&
Autom
at
ion,
(pp
.
1272
~ 1275).
Qingdao,
Ch
ina
:
Air
Forc
e
Avia
t
ion
Univer
si
t
y
.
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
Th
e c
onditi
on
monitori
ng o
f
diesel
engine
s
us
in
g ac
ou
sti
c
signal
analysis
(
Wi
di Praset
yo
)
185
[10]
Hocine
Bendjam
a,
Sala
h
Bouhou
che
,
&
Moham
e
d
Seghir
Bouche
rit
.
(2012).
Appl
ic
a
ti
on
of
W
ave
l
et
Tra
nsform
for
Fault
Di
agnosis
in
Rot
at
ing
Mec
hine
r
y
.
Int
ern
ati
onal
Journa
l
of
Mac
hine L
ea
rni
ng
and
Com
puting,
2
(1), 82
~ 8
7.
[11]
Cao
Shuhua,
Xu
Junjun,
&
Ning
Da
y
ong.
(2015)
.
Air
Valve
Clea
ran
ce
Faul
t
Diag
nosis
of
Diese
l
Engi
ne
Based
o
n
Acoustic
Signa
l
Data
Proc
essing.
Int
ern
ational
Confer
ence
on
Mec
hat
ron
ic
s,
E
le
c
troni
cs,
Industria
l
and
Con
tro
l
Engi
ne
eri
ng
(M
EIC),
(pp
.
1
256
~ 1259).
D
al
i
an,
China
:
Dalian
M
ari
ti
m
e
Univer
sit
y
.
[12]
Qiang
W
ang,
Xuem
in
Ti
an.
(20
13).
Soft
Sensin
g
Based
on
Hilb
ert
-
Huang
Tra
ns
form
and
W
ave
l
et
Support
Vec
t
or
Mac
hine. TEL
K
OM
NIK
A
Indone
sian
Journa
l
of
El
e
ct
ri
ca
l
Eng
i
n
ee
ring
,
11
(7), 3
704
~ 3710.
[13]
H
Li
,
P
Zhou
,
&
X
Ma.
(2005
).
Pat
ern
Re
cog
nit
ion
on
Di
ese
l
Engi
ne
W
orking
Condit
ion
b
y
Us
ing
a
Nove
l
Methodol
og
y
–
Hilbe
rt
Spect
r
u
m
Ent
rop
y
.
Journal
of
Marin
e En
gine
er
ing
&
T
echolog
y
,
4
(1), 43
-
48.
[14]
Micha
e
l
Feldma
n.
(2011)
.
Hilb
er
t
Tr
ansform
in
Vibra
ti
on
Anal
y
si
s.
Mec
han
ical
S
y
stems
and
Signal
Proc
essing,
2
5
(3),
735
-
802
.
[15]
Micha
e
l
Feldman.
(2006).
Ti
m
e
-
Var
y
ing
Vibr
at
i
on
Dec
om
positi
on
and
Anal
y
sis
Based
on
the
Hilbe
rt
Tra
nsform
.
Journal
of
Soun
d
and
Vibr
ation,
295
(3), 518
-
530
.
[16]
Steve
n
W
.
Sm
it
h.
(1997).
Th
e
Scie
n
ti
st
and
Eng
ine
er’
s
Guide
to
Digit
al
Signa
l
Proce
ss
ing.
Cal
i
fornia
:
Ca
li
forn
i
a
Te
chn
ic
a
l
Publis
hing.
[17]
Dr.
M.
J.
de
Sm
it
h.
(2015)
.
St
at
is
ti
c
al
Ana
l
y
s
is H
andbook.
Uni
te
d
Kingdom
:
The
W
inc
hel
sea
Pres
s
[18]
Klaus
Molle
nh
a
uer
,
H
el
m
ut
Ts
c
hoeke
.
(2010). H
andbook
of
D
ie
s
el
Engi
ne
.
Lond
on:
Springer
.
Evaluation Warning : The document was created with Spire.PDF for Python.