TELKOM
NIKA
, Vol. 11, No. 9, September 20
13, pp.
5322
~53
2
8
ISSN: 2302-4
046
5322
Re
cei
v
ed Ma
rch 1
2
, 2013;
Re
vised June
11, 2013; Accepte
d
Ju
ne
21, 2013
Application of Empirical Mode Decomposition for
Ultrasonic Test
ing of Coarse-grained Materials
Qiufeng Li*
1
,
Gengsh
e
ng
Luo
2
, Guo Chen
1
, Pan Huang
1
1
Ke
y
L
abor
ator
y of No
ndestru
c
tive T
e
sting (Nanc
han
g H
a
n
g
kon
g
Univ
ersi
t
y
), Ministr
y
of Educati
on,
Nanc
han
g 33
0
063, Jia
n
g
x
i, Chin
a
2
Colle
ge of Me
chan
ical a
nd El
ectrical En
gin
e
e
rin
g
, Central
South Un
iversit
y
, Ch
angs
ha 4
100
83, Hu
nan,
Chin
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: qiufen
gl
ee@
163.com
A
b
st
r
a
ct
In ultras
onic
te
sting
of co
arse
-grain
ed
mater
i
als,
sig
n
a
l
to no
i
s
e ra
ti
o (SNR
) o
f
te
stin
g sig
n
a
l
s
is
reduc
ed ser
i
o
u
s
ly for the stru
cture no
ise,
a
n
d
ech
oes fro
m
defects are
diff
icult to b
e
i
den
tified. In ord
e
r
to
improve t
he S
NR an
d the r
e
lia
bi
lity of
ult
r
ason
ic testing
of coarse-
g
ra
ine
d
materi
als,
empiric
a
l
mo
d
e
deco
m
positi
on
(EMD) is i
n
trod
uced
to pr
oces
s the te
sti
ng s
i
gna
l h
e
re. Si
g
nal
env
elo
p
e
c
an
be for
m
ed
by
usin
g cub
i
c spl
i
ne i
n
terp
olati
o
n, and n
o
n
lin
e
a
r and
non-
st
ation
a
ry sign
al c
an be
dec
omp
o
sed se
lf-ad
a
p
t
ive
into the sum
of
several intrinsic m
o
de
functio
n
s (IMF
) by using char
acterist
i
c
time scal
e
of the sign
als, an
d
then
hig
her
ord
e
r a
nd te
nd
enc
y of the
ori
g
in
a
l
sig
n
a
l
s ca
n b
e
o
b
tain
ed. T
h
e d
eno
isin
g
ex
peri
m
e
n
t w
i
th l
o
w
SNR si
mu
late
d sign
al
are a
c
hiev
ed acc
o
r
d
in
g to
the fe
ature of EMD,
and SN
R is
enh
anc
ed
mor
e
by
comparis
on w
i
th the w
a
velet ana
lysis metho
d
. And test
ing
sign
al col
l
ecte
d from coars
e
-
g
rai
ned
mater
i
als
is use
d
to fi
nis
h
d
eno
isin
g
ex
peri
m
e
n
t, an
d i
t
is s
how
n fro
m
th
e ex
per
iment res
u
lt that t
he EMD
h
a
s b
e
tte
r
ada
ptive
abi
lity
in d
e
co
mp
osi
ng n
o
ise-
po
llut
ed si
gna
ls
a
n
d
less e
m
piric
a
l
infor
m
ati
on is
requ
ired
in th
e
den
oisi
ng pr
oc
ess.
Ke
y
w
ords
:
ultr
ason
ic testing,
coarse-
g
rai
n
e
d
materi
als, sig
n
a
l de
nois
i
n
g
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Coa
r
se-grai
n
ed mate
rial
s have
som
e
advanta
g
e
s
, such a
s
anti-corro
s
i
on, high
temperature
cre
ep ability, good low-tempe
r
ature
tough
ne
ss. In rece
nt
years, this kind
of
material
s
wa
s u
s
e
d
mo
re
widely i
n
mo
dern
ind
u
st
ry, for
example
the chemi
c
al
rea
c
tion ve
ssels
and
pipin
g
, li
quid
and
ga
s
storage
an
d tra
n
spor
tati
on, nu
cle
a
r
power an
d o
t
hers in
du
stri
al
depa
rtment
s [1, 2]. Therefore, it is esse
ntial for
this kind of materi
als to non
-de
s
tru
c
tive testi
n
g
(NDT) b
o
th in
the produ
ctio
n pro
c
e
ss a
n
d
in the daily maintena
nce.
Ultra
s
oni
c d
e
tection te
ch
nology ha
s
many ch
ara
c
teristics, such as g
r
e
a
t detectin
g
depth, high
sensitivity, gre
a
t penetratin
g
power,
po
sitioning a
c
curacy, lowe
r co
st, high spee
d
,
harml
ess to h
u
man b
ody a
nd ea
sy to field use [3, 4], and so ultra
s
onic te
sting t
e
ch
nolo
g
y is the
most wid
e
ly use
d
than others ND
T m
e
thod
s. Duri
n
g
the Ultra
s
o
n
ic testin
g of coarse
-g
rain
ed
material
s, strongly ultra
s
o
n
ic scatteri
ng
is pr
od
uced
on interfa
c
e b
e
twee
n those
irreg
u
lar a
n
d
big
grain
s
, which
gives ri
se t
o
se
riou
s
structural
noi
se
and ultraso
n
ic en
ergy attenuation, a
n
d
besi
d
e
s
ran
d
o
m noise from aco
u
sti
c
-electri
c st
rin
g
s mixed be
tween the transmitting a
nd
receiving tran
sdu
c
e
r
an
d test sy
stem v
a
riou
s,
which
lead to d
e
crease dete
c
tin
g
se
nsitivity and
defect
s
d
e
tection rate
of [5
-6]. The
rese
arch fo
cal
poi
nt is
ho
w to
restrai
n
stro
ng
noi
se
of si
gn
als
and imp
r
ove the sig
nal-to
-
n
o
ise ratio (SNR) an
d defe
c
ts dete
c
tion ra
te.
From th
e pe
rspe
ctive of di
gital sig
nal p
r
oce
s
sing, the
r
e a
r
e m
any
comm
on a
p
p
r
oa
che
s
for enha
nci
n
g
the SNR of the co
arse
-grained mate
ri
a
l
s dete
c
ting si
gnal at pre
s
e
n
t, such a
s
split
spe
c
tru
m
a
n
a
lysis techno
logy, averag
e filter te
ch
n
o
logy a
nd
correlation
an
alysis, f
r
eq
ue
ncy
spe
c
tru
m
ana
lysis an
d wavelet analysi
s
t
e
ch
nolo
g
y,
by which te
ch
n
o
logie
s
de
sirable results h
ad
be obtain
ed i
n
som
e
re
sp
ects a
nd so
me wea
k
p
o
ints we
re a
ssignabl
e [7-10
]. The empiri
cal
mode de
com
positio
n (EM
D
) was first prop
os
ed in
1998 by Hu
ang, whi
c
h
wa
s suitabl
e
for
pro
c
e
ssi
ng n
on-lin
ea
r and
non-station
a
r
y sign
als [1
1]. EMD is si
milar to the
wavelet an
al
ysis,
but the ba
sis function i
s
n
o
t neede
d to
be defin
e
d
, and the time
scale can b
e
automati
c
al
ly
adju
s
ted with
the local fea
t
ures of the detecti
n
g
data by cubic
spline interpol
ation, and si
gnal
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Applicatio
n of Em
pirical Mo
de De
com
p
o
s
ition for Ult
r
aso
n
ic Te
stin
g of… (Qiufe
ng Li)
5323
details
are
g
o
tten layer
b
y
layer, and
the ori
g
i
nal
signal
can
be
rep
r
e
s
ente
d
by a serie
s
of
narro
w-b
and
stationa
ry i
n
trinsi
c m
o
d
e
func
tio
n
s
(IMF) an
d a
resi
dual, a
n
d
therefo
r
e hi
gh
freque
ncy
re
solution can b
e
gotten. By comp
ari
s
o
n
with othe
rs m
e
thod
s, EMD
is ad
aptive time-
freque
ncy
an
alysis
metho
d
witho
u
t an
y prior
kn
owl
edge [1
2-1
4
]. In this
contri
bution, be
ca
use
stru
ctural n
o
i
s
e
of ultra
s
o
n
ic te
sting
si
gnal of
th
e
coarse
-graine
d
mate
rials i
s
n
on-li
nea
r,
non-
stationa
ry, EMD method i
s
applie
d to decompo
se t
he detectin
g
sign
al and IMF comp
one
nts
forming th
e d
e
tecting
sig
n
a
l are obtai
n
ed at fi
rst, an
d then
stru
ct
ural
noi
se a
n
d
environme
n
tal
noise can b
e
remove
d by sifting IMF co
mpone
nts,
an
d the SNR a
nd defe
c
ts
de
tection
rate
can
be improved.
2.EMD Ba
sic
Principles
EMD is fo
rm
ed ideol
ogi
ca
lly acco
rdi
ng
to
the though
t which any signal
s are co
mposed
by the dif
f
erent intrinsic vibration mo
de
s.
And EMD is a process of
smoothing
the signal
x(
t
)
in
essen
c
e:
The
delay betwe
en the
adjacent peak poi
nt of
the decom
po
sed si
g
nal is defined
as
the time scal
e at first, an
d then a
seri
es of
station
a
ry si
gnal
s
with dif
f
erent
time scale
s
c
i
(t)
(
i
=1
…n
) and
resi
dual
R(t)
can be
cal
c
ul
ated as follo
wing eq
uatio
n after the signal
s are sift
ed
and de
com
p
o
s
ed.
1
()
()
()
n
i
i
x
tc
t
r
t
(1)
Her
e
c
i
(t)
de
notes dif
f
e
r
e
n
t orde
r IMF
comp
onent
respe
c
tively whi
c
h mu
st meet two
con
d
ition
s
: Firstly
,
wheth
e
r the number
of t
he extrema point and the ze
ro cro
s
sing poi
nt in
h(t)
are eq
ual or
dif
f
er by at most one.
And
se
co
n
d
ly
, average valu
e of two envelop
es co
mpo
s
e
d
b
y
any point, th
e local maxima point and local mini
ma point must be zero.
The whol
e si
fting
flowchart is shown in Figure 1,
these steady IMF compone
nts refl
e
c
t vibration mode
s from high
freque
ncy to
low fre
que
ncy in accorda
n
ce
with the
decompo
sitio
n
se
que
nce resp
ectively
, a
n
d
the remai
n
ing
resid
ual refle
c
ts the ove
r
al
l trend of the origin
al sig
nal
.
Figure 1. Flowchart of EM
D
T
e
rminatio
n condition
s of whol
e sifting pro
c
e
ss
will af
fect the deco
m
positio
n ef
fect, and
so jud
g
ment
con
d
ition
s
for termination
commonly are
that EMD decomp
o
sitio
n
is finish
ed wh
en
the remaind
e
r
R
n
(t)
is a monotoni
c function or ener
g
y
of the last
decompo
se
d
c
n
(t)
or
R
n
(t
)
is
lowe
r pre-set
value (Pvalue) a
c
cordin
g to proje
c
t need
s [1
1]. It can be ju
dg
ed as follo
wi
ng
equatio
n wh
e
t
her the latter conditio
n
is
met.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 9, September 201
3: 532
2 – 5328
5324
Pvalue
)
(
)
(
)
(
N
2
0
1
2
0
1
T
t
k
T
t
k
k
t
h
t
h
t
h
(2
)
Her
e
h
k
(t
)
an
d
h
k-1
(t)
a
r
e th
e re
sidu
al si
g
nals
after the
k
th an
d
(k
-1)
t
h
sifting, an
d
Pvalue expre
s
s
the pre
-
set value.
EMD also ha
s well
spatio
-tempo
ral filteri
ng
ability
,
and
rebuildi
ng si
gnal can be o
b
tained
by selectin
g dif
f
erent IMF compo
nent
s accordi
ng to
requireme
nt and low-pa
ss filtering, hi
gh-
pass filter an
d band
-pa
s
s filtering a
r
e ab
le to be achi
e
v
ed respe
c
tively
.
3. Simulation Rese
arch
A
simulated sign
al is design
ed as Fi
gure
2(a) a
c
cording to the actually detecte
d
con
d
ition
s
, which in
clud
es,
the defect wave and the bottom wave,
and the dete
c
ting freq
uen
cy
of the sign
al is 2.5M
Hz.
An
d then the o
r
i
g
inal si
gnal
x (
t
)
with 2dB
SNR i
s
obtai
ned a
s
sho
w
n in
Fig. 2(b) afte
r rand
om noi
se is add
ed into. From
the figure, the reflecte
d wav
e
from defect
is
subm
erged in
to the noise,
and me
rely the starting
wa
ve and the bo
ttom wave ca
n be se
en.
Figure 2. Simulated Te
st Signal
(a)
(b)
Figure 3. Wa
veform and S
pectrum of T
e
st Signal an
d Its IMF
0
2
4
6
x 1
0
-5
-1
-0
.
5
0
0.
5
1
t / s
M
agni
t
ude
(b)
0
2
4
6
x 1
0
-5
-1
-0
.5
0
0.
5
1
t / s
M
agni
t
ude
(a)
-1
0
1
x(
t
)
-0.
2
0
0.
2
-0.
1
0
0.
1
-1
0
1
-0.
5
0
0.
5
-0.
2
0
0.
2
-0.
0
5
0
0.
05
-0.
0
2
0
0.
02
-0.
0
2
0
0.
02
-0.
0
1
0
0.
01
-5
0
5
x 1
0
-3
-5
0
5
x 1
0
-3
0
1
2
3
4
5
x 1
0
-5
-4
-2
0
x 1
0
-3
t /
s
IM
F1
IM
F2
IM
F3
IM
F4
IM
F5
IM
F6
IM
F7
IM
F8
IM
F9
IM
F10
IM
F1
1
R
M
a
gni
t
u
de
0
50
100
0
50
0
50
0
50
100
0
50
0
50
0
50
0
50
0
50
0
5
10
0
5
10
0
5
10
0
1
2
3
4
5
x 1
0
7
0
5
10
f /
H
z
M
a
gni
t
ude
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Applicatio
n of Em
pirical Mo
de De
com
p
o
s
ition for Ult
r
aso
n
ic Te
stin
g of… (Qiufe
ng Li)
5325
After the sign
al with noise is de
comp
osed ba
sed on
EMD method
, 11 IMF com
pone
nts
and a re
sid
u
a
l comp
one
n
t
R are obtai
ned. The tim
e
-do
m
ain wa
veforms a
n
d
corre
s
po
ndi
n
g
s
p
ec
tr
ums
o
f
x(t
)
and th
e IMF co
mpo
n
e
n
ts a
r
e li
sted
in Figu
re 3. It can
be fou
n
d
from the
who
l
e
decompo
sitio
n
pro
c
e
s
s th
at bigge
r the
IMF or
d
e
r
and
smalle
r
the ce
ntral freque
ncy of t
he
corre
s
p
ondin
g
IMF comp
onent, whi
c
h
is similar to
the conditio
n
of wavelet decompo
sin
g
the
sign
al, and
h
o
weve
r the freque
ncy b
a
n
d
of differ
ent I
M
F co
mpon
e
n
t decompo
sed by EMD i
s
not
fixed and
but
ch
ang
ed
wit
h
ad
aptive v
a
riation
of
E
M
D
based
o
n
the
station
a
ry of th
e
si
gnal.
From Fig
u
re
3, the first
two IMF co
mpone
nts ha
ve large
s
t freque
ncy sp
e
c
trum rang
e and
smalle
r am
pli
t
ude which
a
r
e cha
r
a
c
teri
stics of
the
noise, and
can be ju
dge
d that the m
a
jor
comp
one
nts of
two comp
onent
s
a
r
e noise.
And
however
wa
ve packet i
s
obviou
s
in
the
remai
n
ing IM
F comp
one
nts sta
r
ting fro
m
3
th ord
e
r I
M
F whi
c
h a
r
e useful
com
pone
nts. And
so
the first two
orde
r IMF
co
mpone
nts directly remove
d an
d th
e o
r
i
g
inal
sig
nal i
s
reb
u
ilt with
the
remai
n
ing
IMF compo
nent
s a
nd th
e
re
sidu
al. Thi
s
method
of re
building
sign
al is si
mple
r
and
more intuitive unlike
wav
e
let deco
m
p
o
sition which
is complex
and nee
ded
to calcul
ate the
rebuil
d
ing co
efficient.
The co
mpa
r
i
s
on of EMD denoi
sing result an
d wa
velet denoi
sing re
sult is
sho
w
n in
Figure 4, and
the wavefo
rm 1 is the
ori
g
inal
signal
, t
he waveform
2 is the
sign
a
l
with noi
se, the
waveform 3 is re
built sig
n
a
l after EMD
denoi
sing
an
d the wavefo
rm 4 is the re
sult with
wav
e
let
denoi
sing
wh
ich is a
c
hiev
ed with db4
wavelet and t
he thre
shol
d automatica
lly sele
cted. From
the figure, de
fects wave ca
n be ef
fective
l
y found by
both re
sults, a
nd but in the
result of wav
e
let
denoi
sing, wa
ve packet is not smooth a
nd many
usef
ul compo
nent
s are re
move
d.
And howev
er
EMD metho
d
is appli
ed to
denoi
se, ba
sis fu
nction
n
eed not to b
e
sele
cted
a
nd ca
n ad
apt
ive
obtain d
a
ta e
n
velope
and
the
ef
fective
ingre
d
ient
s
are remaine
d
well, and the
filtering re
sult can
be better talli
ed with the
origin
al sig
n
a
l
.
The comp
arison with b
o
th filtering result
s detail
s
is
achi
eved a
s
sho
w
n in
Fig
u
re 5, a
nd th
e re
sult
s
of EMD den
oisi
ng metho
d
is more
con
s
i
s
tent
with the origi
nal signal fro
m
the figure.
The
SNR of both results can be cal
c
ulate
d
by th
e
Equation 3 [1
5, 16].
2
1
lg
10
SNR
P
P
(3)
Her
e
P
1
is th
e ene
rgy of t
he o
r
iginal
si
gnal, an
d
P
2
expre
s
ses th
e ene
rgy of
si
gnal afte
r
denoi
sing.
After cal
c
ulatin
g
,
the
SNR after W
a
vel
e
t denoi
sing is 4.62dB and the
SNR after EMD
denoi
sing i
s
1
1
.01dB.
4. Experimental Tes
t
Experimental
spe
c
ime
n
is sho
w
n i
n
Fi
g. 6, it
is a cast iro
n
with
2-level g
r
ain
degree.
Thickne
s
s of
the sp
eci
m
en
is 6
0
mm, an
d a flat botto
m hole
whi
c
h
ape
rture
is
Φ
2mm is located
Figure 4. Co
mpari
s
o
n
of Den
o
isi
ng Re
sult by
EMD and
Wa
velet
Figure 5. Co
mpari
s
o
n
of Den
o
isi
ng De
tail by
EMD and
Wa
velet
0
1
2
3
4
5
x 1
0
-5
-3
-2
.5
-2
-1
.5
-1
-0
.5
0
0.
5
1
1.
5
2
t /
s
magn
i
t
u
d
e
W
a
v
e
f
o
rm (1
)
W
a
v
e
f
o
rm (2
)
W
a
v
e
f
o
rm (
3
)
W
a
v
e
f
o
rm (4
)
2.
5
8
2.
6
2.
6
2
2.
6
4
2.
6
6
2.
6
8
2.
7
2.
72
x 1
0
-5
-0
.
2
-0
.
1
5
-0
.
1
-0
.
0
5
0
0.
0
5
0.
1
0.
1
5
0.
2
t / s
A
m
pl
i
t
ud
e
/
V
x(
t
)
by E
M
D
by w
a
v
e
l
e
t
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 9, September 201
3: 532
2 – 5328
5326
insid
e
45m
m deep.
The experime
n
t
is completed with ultraso
n
ic pul
se
-echo metho
d
.
Experimental
detection sy
stem is com
p
ose
d
of t
he
Olympus 50
7
7
ultraso
n
ic signal tran
smit
ting
and re
ceiving
instrume
nt, 2C2
0
N ultrasonic tra
n
sd
u
c
er whi
c
h is p
r
odu
ce
d by Shantou Institu
t
e
of Ultra
s
oni
c
Instrum
ents
Co., Ltd., 981
2 data a
c
qui
sition ca
rd a
n
d
comp
uter
. E
x
perime
n
tal test
sign
al is sho
w
n in Figure 7, becau
se th
e test
signal is interfered b
y
seriou
s noi
se and reflect
ed
energy from the flat bottom hole is we
ak, the refl
ect
ed wave from
the flat bottom hole is alm
o
st
subm
erged b
y
noise an
d b
u
t merely the begin
n
ing wa
ve and the bo
ttom wave ca
n be se
en.
Figure 6. Outside Vie
w
of Testing Sp
eci
m
en
Figure 7. Experime
n
tal Testing Signal
After the detecting sig
nal is decom
po
sed
wi
th EMD, 14 IMF compo
nents an
d a resid
ual
can b
e
obtai
ned.
The
wav
e
form an
d fre
quen
cy sp
ect
r
um of the IM
F com
pone
nts are sho
w
n i
n
Figure 8. Fro
m
the figure,
the first 4 IMF com
pon
e
n
ts have la
rg
est freq
uen
cy
spe
c
trum
ra
nge
and smalle
r amplitude an
d
so sh
ould
be
re
move
d
as n
o
ise, a
nd ho
weve
r
the others I
M
F
comp
one
nts starting
from the
5
th o
r
de
r
IMF have hig
h
sp
ect
r
um e
nergy a
nd
sm
all ban
ds
whi
c
h
are u
s
eful
co
mpone
nts of
sign
als.
And so
the sig
nal can be rebuilt
by
the
IMF
compon
ents after
the
5
th order
, and the compari
s
o
n
of the rebuil
d
ing
signal and the origin
al si
gnal is sho
w
n in
Figure 9. It can be found that t
he reflected signal fro
m
the flat
bottom hole has stood out from
stron
g
noi
se
and the refle
c
tion positio
n is so o
b
viou
s from the figure.
Figure 8. Wa
veform and S
pectrum of T
e
st Signal an
d Its IMF
0
0.
5
1
1.
5
2
2.
5
x 1
0
-5
-0
.
6
-0
.
4
-0
.
2
0
0.
2
0.
4
0.
6
t /
s
A
mp
lit
u
d
e
/
V
B
o
t
t
om
w
a
ve
S
t
ar
ti
ng wa
ve
-0
.
5
0
0.
5
IM
F
1
-0
.
5
0
0.
5
IM
F2
-0
.
5
0
0.
5
IM
F3
-0
.
5
0
0.
5
IM
F4
-1
0
1
IM
F5
-0
.
5
0
0.
5
IM
F
6
-1
0
1
IM
F
7
-0
.
5
0
0.
5
IM
F
8
-0
.
2
0
0.
2
IM
F9
-0
.
1
0
0.
1
IM
F
1
0
-0
.
1
0
0.
1
IM
F
1
1
-0
.
0
5
0
0.
0
5
IM
F12
-0
.
0
2
0
0.
0
2
IM
F13
-0
.
0
1
0
0.
0
1
IM
F14
0
1
2
3
4
x 1
0
-5
-0
.
0
1
0
0.
0
1
t / s
R
M
a
gni
t
ude
0
50
0
50
0
50
0
50
0
50
10
0
0
50
10
0
0
10
0
20
0
0
10
0
20
0
0
10
0
20
0
0
50
10
0
0
50
10
0
0
50
0
50
0
50
0
2
4
6
8
10
12
x 1
0
7
0
50
f / H
z
M
agn
i
t
ud
e
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Applicatio
n of Em
pirical Mo
de De
com
p
o
s
ition for Ult
r
aso
n
ic Te
stin
g of… (Qiufe
ng Li)
5327
Figure 9. Co
mpari
s
o
n
of Wavefo
rm
s b
e
fore an
d after EMD Processing
5. Conclusio
n
In the
cont
rib
u
tion, EMD
method
whi
c
h was ap
plie
d to p
r
o
c
e
s
s
the dete
c
ting
sig
nal fo
r
stron
g
stru
ct
ural
noi
se
was
produ
ce
d
in
ultra
s
oni
c
t
e
st
of
c
o
a
r
se-
g
r
a
in
mat
e
rial
s.
A
n
d
t
i
me
domain
an
d f
r
equ
en
cy
spe
c
trum
of IMF
co
mpo
nent
s we
re
gotten
by EMD,
and
then
the
noi
se
sign
al was
sep
a
rate
d from the te
st
sig
nal a
n
d
effective
se
ction
s
were
picked
out.
The
comp
ari
s
o
n
with de
noi
sin
g
effect b
o
th
by EMD a
n
d
wavel
e
t wa
s a
c
hieve
d
thoug
h num
erical
simulatio
n
, a
nd it could
b
e
sh
own fro
m
the co
mp
a
r
iso
n
results
that the EMD method fo
cu
se
s
more
on
the
intrinsi
c mod
e
comp
onent
s a
n
d
co
uld
be a
daptive
sign
al d
e
com
positio
n
with
out
any prio
ri info
rmation of th
e origi
nal si
g
nal, and
b
e
tter filtering
effect could b
e
obtaine
d after the
first ord
e
r I
M
F com
pon
ents which have the no
is
e
ch
ar
ac
te
ris
t
ics
w
e
r
e
re
mo
ve
d
.
F
i
na
lly,
decompo
sitio
n
and
sig
nal
rebuil
d
ing fo
r the expe
rim
ent sig
nal
we
re a
c
hieve
d
, and it could
be
sho
w
n from t
he experim
en
t results that the re
sult
pro
c
essed by EMD wa
s more effective to filter
out the stro
ng
noise
sign
al, and the re
fle
c
ted sign
al co
uld be seen o
b
viously.
Ackn
o
w
l
e
dg
ments
This wo
rk wa
s sup
port
ed
by Nati
onal Natu
ral
Scien
c
e
F
ound
ation
of
Chi
na
(112
640
32
,
1
1104
129
), by Aeron
autical
Scien
c
e Fo
u
ndation of
China(201
1ZE
5600
6), Natu
ral
Science F
o
undation of
Jiangxi Province
(20122BAB201024), by Graduate Innovation
Found
ation o
f
Nanchan
g Han
g
kong
University
(Y
C2012
012
) an
d by the Gra
duate Innova
t
ion
B
a
se of
Jia
n
g
x
i P
r
ov
ince.
Referen
ces
[1] Alava,
Mikko.
Coarse-
g
rai
n
ed
materi
als
properti
es for
fiber-b
ase
d
mater
i
als fr
om c
o
mpute
r
simulati
ons.
Pr
ocee
din
g
s
of Internati
o
n
a
l
C
onfere
n
ce
on
Nan
o
tech
nol
og
y for
the
F
o
re
st Product
s
Industr
y
.
Esp
o
o
. 2010; 6
10-6
33.
[2]
Bingfa
ng W
A
N
G, Z
andong
H
A
N, Ke
yi YUA
N
.
Sign
al pr
oc
essin
g
in
ultra
s
onic test of a
u
stenitic w
e
l
d
s
base
d
on ti
me-
f
reque
ncy an
al
ysis.
T
r
ansactions Of
T
he Chi
n
a
w
e
l
d
i
ng Insti
t
ution. 20
11; 3
2
(5): 25-2
9
.
[3]
BS Ben, B
A
Ben, C
h
R
a
tn
am. Ultras
onic
bas
ed
m
e
tho
d
for
dama
g
e
id
entificati
o
n
in c
o
mpos
it
e
materials.
Inter
natio
nal J
ourn
a
l of Mecha
n
ic
s and Materi
als
in Desi
gn
. 20
1
2
; 8(4): 297-3
0
9
.
[4]
Yang J
i
n
hai, T
ohi
di, Ba
hman.
Char
acterizati
on of
inh
i
biti
on
mecha
n
isms
of kinetic
h
y
dr
ate in
hi
bito
r
s
usin
g ultraso
n
i
c
test techniqu
e
. Che
m
ic
al En
gin
eeri
ng Sci
e
nce
. 201
1; 66(
3): 278-2
83.
[5]
Lin
g
Xian
g, Ma Gang. Effect of ultrasonic i
m
pact treatme
nt on the stress corrosio
n
cra
cking of 30
4
stainl
ess
steel
w
e
ld
ed joi
n
ts.
Journ
a
l of Pr
essure V
e
ssel
T
e
chno
logy,
T
r
ansactio
n
s o
f
the ASME
.
200
9; 131(
5): 0515
02
1-05
15
0
25.
[6]
Matz Vacl
av, S
m
id R
adis
l
av,
Starman Sta
n
i
s
lav.
Sig
n
a
l
-to-
nois
e
rati
o e
n
h
ance
m
ent b
a
s
ed
on w
a
ve
let
filterin
g in ultra
s
onic testin
g
. Ultraso
nics. 20
09; 49(8): 7
52-
759.
[7]
Z
hu Yo
ng, W
e
ight, Jo
hn
P.
Ultraso
nic
non
destructive
ev
alu
a
tion
of
hig
h
l
y
sc
attering
materials
us
in
g
ada
ptive filteri
ng an
d detecti
on.
IEEE Transactions o
n
Ul
trasonics, Ferroel
ectrics, and
Frequen
c
y
Contro
l
. 199
4; 41(1): 26-
33.
[8]
LIU Z
hen
qin
g
,
Ll Ch
en
gli
n
, W
E
I Moan.
F
l
aw
-to-grain echo enh
anc
e
m
e
n
t
by
cros
s-correlati
o
n
w
e
ightin
g an
d split-sp
e
ctru
m process
i
ng
. AC
T
A
ACUST
I
CA. 1996; 21(
4): 714-7
26.
0
0.
5
1
1.
5
2
2.
5
x 1
0
-5
-0
.5
0
0.
5
1
1.
5
t / s
M
agni
t
ude
R
e
f
l
e
c
t
e
d
w
a
v
e
fro
m
flat-
bottom
hole
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 9, September 201
3: 532
2 – 5328
5328
[9]
Drai
R, Benammar A, Benchaala A.
S
i
g
nal
proc
essin
g
for the
detec
tion of
multip
l
e
i
m
p
e
rfectio
n
echo
es drow
ne
d in the structu
r
al no
ise.
Ultra
sonics. 20
04; 4
2
(1-9): 83
1-83
5.
[10]
He Min
gge, Yi
n Ying, L
i
n L
i
j
un. Den
o
is
ing
of Ultrason
i
c T
e
sting
Si
gn
al
Based o
n
W
a
v
e
let F
ootpri
n
t
and Match
i
n
g
Pursuit.
Journ
a
l
of basic scie
n
c
e and e
n
g
i
ne
erin
g
. 201
1; 19
(2): 297-3
04.
[11]
NE H
uan
g, Sh
en Z
,
L
o
n
g
S
R
.
T
he e
m
piric
a
l
mod
e
deco
m
p
o
sitio
n
a
n
d
Hi
l
bert sp
ectru
m
f
o
r n
onl
in
ear
and
no
n-statio
nary ti
me s
e
ri
es an
alysis
. In
the Proc
eed
in
gs of the
Ro
yal Soc
i
et
y A. L
ond
on. 1
9
9
8
;
903-
995.
[12]
Hua
ng Nor
den
E, W
u
Man-Li,
Qu W
endon
g.
Appl
icatio
ns of Hilb
ert-Hu
ang t
r
ansfor
m
to no
n-station
a
ry
financ
ial ti
me s
e
ries
ana
lysis
.
Appl
ied Stoc
ha
stic Mode
ls in
Bu
sin
e
ss an
d I
ndustr
y. 20
03;
19(3): 2
45-
268.
[13]
Gai Guan
gho
n
g
. T
he processing of rotor s
t
art
up sig
nals
base
d
on
emp
i
rical m
ode
de
compos
ition
.
Mecha
n
ica
l
Systems a
nd Sig
nal Proc
essin
g
. 2006; 20(
1): 222-2
35.
[14]
Hao
Din
g, Z
h
i
y
a
o
H
u
a
ng, Z
h
ih
uan
Son
g
.
Hilb
ert-Hu
ang
transform bas
ed si
gn
al a
n
a
l
ysis for th
e
character
i
z
a
tio
n
of
gas-l
iq
uid
tw
o-phase
flo
w
. F
l
o
w
M
eas
ureme
n
t a
nd I
n
strumentati
o
n
.
200
7; 1
8
(1
)
:
37-4
6
.
[15]
Kim J, Ud
pa
L, Udp
a
S. M
u
lti
-stag
e
a
dap
tive no
ise c
a
n
c
ellati
on
for u
l
trasonic
NDE.
NDT
and
E
Internatio
na
l. 2001; 34(
5): 319
-328.
[16]
Yang L
e
, He Xiao
xi
an
g, Yang
Gang. A nove
l
blin
d eq
ual
iza
t
ion metho
d
for recombin
atio
n of signa
ls
.
Internatio
na
l Journ
a
l of Dig
ita
l
C
onte
n
t T
e
chnol
ogy a
nd its Appl
icatio
ns
. 2
011; 5(8): 2
98-
307.
Evaluation Warning : The document was created with Spire.PDF for Python.