TELKOM
NIKA
, Vol.13, No
.1, March 2
0
1
5
, pp. 65~7
5
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v13i1.1320
65
Re
cei
v
ed O
c
t
ober 1
3
, 201
4; Revi
se
d Decem
b
e
r
14, 2014; Accept
ed Ja
nua
ry 1
2
, 2014
Ultrasonic Tomography of Immersion Circular Array by
Hyperbola Algorithm
Liu Yang
1
*, Chung
uang
Xu
1
, Xianghui Guo
1
, Lipin
g
Wang
2
1
Ke
y
La
bor
ato
r
y
of F
und
ame
n
tal Scie
nce fo
r Advance
d
Ma
chini
ng, Bei
j
i
n
g
Institute of
T
e
chno
log
y
5 Zhong
gu
anc
un South Stre
e
t, Haidia
n Distri
c
t, Beijing, 10
0
081, Ch
in
a
2
Sansom Institute for Health
Rese
arch an
d
Schoo
l of
Phar
mac
y
a
nd Me
di
cal Scie
nces, Univers
i
t
y
of
South Austral
i
a
,
Adelai
de, So
uth
Au
stra
l
i
a
, 500
1
,
Au
stra
l
i
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: 3120
10
017
9
@
bit.ed
u
.cn
Abs
t
rak
T
h
is pap
er pre
s
ents a dev
elo
p
ment an
d res
earch
of a n
o
n
-inv
asive u
l
trason
ic tomogr
aphy fo
r
imagi
ng
gas/l
iq
uid tw
o-p
hase
flow
. Ultrasoni
c trans
mitting
and r
e
ceiv
in
g
are i
m
ple
m
ent
ed us
in
g a c
i
rcul
a
r
array
mo
de
l t
hat co
nsists
of 36
trans
du
cers. CO
MSOL Multi
phys
i
cs
®
softw
are is ad
opte
d
for
the
simulati
on
of the ultras
on
ic p
r
opa
gatio
n in t
he d
e
tect
ing
zone. Vari
ous t
w
o-phase fl
ow
s w
i
th different gas
distrib
u
tions ar
e radi
ated by u
l
tras
on
ic w
a
ves and the refl
e
c
tion mod
e
ap
proac
h is utili
zed for detecti
n
g
th
e
scattering
w
a
ves after t
he
ge
nerati
on
of fan
-
shap
ed
be
a
m
. Ultras
onic
attenu
atio
n a
n
d
soun
d s
pee
d
a
r
e
both taken i
n
to
consid
eratio
n w
h
ile reco
nstructing
the tw
o-phas
e flow
images u
n
d
e
r the inh
o
m
og
en
e
o
u
s
me
di
um c
ond
itions. T
he inv
e
r
s
ion pr
oced
ure
of the ima
ge
reconstructi
on
is reali
z
e
d
us
i
ng the hyp
e
rb
ola
alg
o
rith
m, w
h
ich in return d
e
m
onstrates the feasi
b
il
ity and v
a
lid
ity of the propos
ed circu
l
ar
array mo
de
l.
Ka
ta
k
unc
i:
Ul
trasonic T
o
mo
grap
hy, Reflect
i
on Mo
de, Hyp
e
rbo
l
a Alg
o
rith
m
1. Introduc
tion
Two
-
ph
ase fl
ow i
s
wide
sprea
d
in
lots of ind
u
stri
al
appli
c
ation
s
,
su
ch
a
s
th
e filling
operation i
n
t
he p
a
int, det
erge
nt a
nd
cosmeti
c
, a
s
well
as the
transportatio
n
of drug
s in
the
pharma
ceuti
c
al indu
stry wh
ere b
ubbl
es
may degrade
the pro
d
u
c
t [1]. Therefore, the dete
c
tion
of
bubbl
es i
s
q
u
i
te necessa
ry
in the
s
e field
s
. Fo
r th
is
re
aso
n
, many t
y
pes of tom
o
grap
hic meth
ods
have bee
n
develop
ed to
measure th
e two-pha
se
flow. Haib
o
Jin et al. use
d
Electri
c
al
Re
sista
n
ce T
o
mog
r
ap
hy (ERT) techniq
ue to i
n
vest
ig
ate the
air-wa
ter two-p
h
a
s
e
flow i
n
a
bu
b
b
le
colum
n
with a
heig
h
t
of 2m
an
d
a di
ameter
of 0.
282m. Saute
r
diam
eters
of bubbl
es were
obtaine
d a
n
d
the lo
cal
axi
a
l velo
city of
the two
-
ph
ase flow was calcul
ated [2].
I. Ismaila
et
al.
pointed
out t
hat Elect
r
ical
Ca
pa
citan
c
e
Tomo
gr
a
phy
(ECT)
co
uld
deal
with
th
e complexity of
multi-ph
ase flow mea
s
u
r
e
m
ent by explicitly
derivi
ng the com
pone
nt distri
bution
s
on two
adja
c
ent plan
es alo
ng a pi
peline [3]. Ro
bert Bana
si
a
k
et al. prese
n
ted a prelim
inary study o
n
automated t
w
o-ph
ase ga
s-l
i
quid flow p
a
ttern ide
n
ti
fica
tion based on
a fuzzy
eval
uation of se
ri
es
of re
con
s
tructed 3
D
ECT
volumetri
c
im
age
s [4]. Th
ese
tomog
r
a
phy metho
d
s we
re
ba
sed
o
n
Electri
c
al Im
peda
nce To
mography
(E
IT), which
d
i
d not h
a
ve
a spatial
re
solution a
s
hi
gh a
s
some oth
e
r
imaging mo
d
a
lities, like
Magneti
c
Re
son
a
n
c
e Ima
g
ing (M
RI), or Ultrasoun
d
Comp
uted To
mography (UCT).
UCT
po
sse
s
ses the
advan
tages i
n
ima
g
ing m
u
lti-ph
ase flo
w
s an
d it is a
b
le to
provi
d
e
quantitative and real-tim
e imagin
g
in
chemi
c
al in
dustry p
r
o
c
e
ss, li
ke filtra
tion [5], without
interrupting t
he process
[6],[7]. In th
e indus
t
rial
proc
es
s
c
onc
erning t
w
o-phas
e
flow, the
approp
riate g
a
s-li
quid
ratio
and
bub
ble
size a
r
e p
r
ov
ed to b
e
two
key fa
ctors.
R.D.M. Carva
l
ho
et al. obtain
e
d
ba
sic ga
s
p
hase st
ru
ctures
and
di
fferent
flow patte
rns
by che
c
ki
ng
the acou
stic
attenuation d
a
ta again
s
t e
x
perime
n
tal data on the
vo
id fractio
n
and flow to
p
o
logy of vertical,
upward, air-water bub
bly flows in the void frac
tio
n
ra
nge from 0 to
15% [8]. M
H F. Rahim
a
n et
al. identified
the flow pa
ttern an
d m
easure
d
th
e
cro
s
s-se
ction
a
l void fracti
on by u
s
in
g
the
ultrasoni
c tra
n
smi
ssi
on mo
de tomog
r
ap
hy (UTT
) [7]-[9]. The two-p
hase flow ima
ges o
b
tained
by
M.H.F. Rahi
man et al.
provide
d
the
informat
ion
of the mixing zon
e
dist
ribution a
n
d
the
comp
one
nt i
n
terface in
complex
sep
a
r
ation
pro
c
e
s
s [10]. Javad
Abba
szade
h
et al. appli
e
d a
non-i
n
vasive
ultraso
n
ic t
e
ch
niqu
e to visuali
z
e the
gas bu
bble
s
usi
ng the
steel pip
e
a
s
a
conveyo
r
an
d
pre
s
ente
d
a
method for vi
suali
z
in
g
the
pipe st
ru
cture
with finite element softwa
r
e
[11]. F.R.M.
Yunus et al.
pro
p
o
s
ed
a
co
mbi
natio
n
of UTT a
n
d
ERT fo
r im
aging
two
-
ph
ase
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 1, March 2
015 : 65 – 75
66
gas/liq
uid flo
w
. The
si
mul
a
tion results
indicated a
g
ood
dete
c
tio
n
of 1
0
-mm
gas bu
bble
s
in a
100-mm diam
eter acrylic v
e
ssel [12].
Although the
r
e has be
en a
rapid devel
o
p
ment in
the field of monitoring the two
-
pha
se
flow, the accurate m
e
asurement is
still
a chall
enge for researchers in
process t
o
mography. The
recon
s
tru
c
ted
image
s are
poor, pa
rticularly in t
he
cente
r
area,
becau
se of il
l-po
sed i
n
verse
probl
em
s a
n
d
limited
me
asu
r
em
ent d
a
ta [12].
Wh
en the
wavel
ength
of the
singl
e ultraso
und
pulse is
com
para
b
le to th
e bub
ble dia
m
eter, the
ref
l
ected wave exhibits com
p
lex
beh
aviors
that
depe
nd o
n
th
e si
ze
and
sh
ape
of the int
e
rface [13].
It may lea
d
to
a blu
rry di
stri
buting im
age
of
multi-bu
bble
s
in the liquid.
In this pape
r, a circ
ula
r
array model
with its se
nsiti
z
i
ng zo
ne at th
e
cente
r
area i
s
firstly esta
b
lishe
d. And then,
36 tra
n
sducers are e
v
enly embed
ded in the wall of
the contain
e
r.
In this way,
an ex
tensive
dataset is pro
v
ided from wh
ich the spati
a
l distrib
u
tion
o
f
the ultraso
n
i
c
p
r
op
agatio
n pa
ram
e
ters
can
be
reco
nstructe
d
usi
ng
a
suitable i
n
version
pro
c
ed
ure. Finally, multi-bubbl
e imag
es of t
he two-ph
ase flow are re
co
nst
r
ucte
d an
d the
accurate po
si
tion and si
ze
of each b
ubbl
e can b
e
dete
r
mine
d.
2. Model and
Method
2.1. Immersion Circular
Array
The
circula
r
array and th
e
config
uratio
n
of t
he propa
gation a
r
e
sketche
d
in Fig
u
re 1. In
this figu
re, 3
6
tran
sdu
c
e
r
s
with 0.5M
Hz f
r
equ
en
cy an
d
5mm di
amet
er a
r
e
evenly
spa
c
e
d
a
r
ou
nd
the circu
m
ference of the
containe
r, in which th
e wate
r is filled
as t
he coupla
n
t. At each m
o
m
ent,
one tra
n
sd
ucer is u
s
e
d
as the transmitter to gen
erat
e a fan-sha
p
ed ultra
s
oni
c radiation. After
th
a
t, a
ll th
e
tr
a
n
s
d
u
c
e
r
s
in
th
e
c
i
rc
u
l
ar
a
r
r
a
y ar
e
u
s
ed
as
th
e r
e
c
e
ive
r
fo
r d
e
t
e
c
ting
and
measuri
ng th
e si
gnal
s i
ndi
vidually. Ultra
s
oni
c
gen
erat
ing a
nd
sig
n
a
l
re
ceiving
a
r
e carried
out
in
seq
uen
ce
so
that the recon
s
tru
c
tion data
can al
so be
colle
cted in o
r
de
r.
Figure 1. Con
f
iguration of u
l
traso
n
ic
p
r
o
p
agation in
sid
e
the circula
r
array.
For the
ga
s/liquid two-p
h
a
s
e flo
w
, the
data obt
ai
ne
d from a
n
ult
r
asoni
c tomo
grap
hy in
the tran
smi
s
sion mod
e
ma
y not able to
be used for
the cha
r
a
c
teristic ima
ge re
con
s
tru
c
tion
of
gas bub
ble
s
whi
c
h
wo
ul
d shado
w th
e sound
wa
ves
due to
a
c
ou
stic impe
dan
ce mi
sm
atch.
Ho
wever, a
n
ultrasoni
c to
mography in
the refle
c
ti
on
mode i
s
real
ized to
be a
solutio
n
for t
h
is
probl
em [14]. This ima
g
ing
pro
c
e
ss i
s
b
a
se
d on a
waveform an
al
ysis in
cludi
ng
arrival time
and
amplitude
cal
c
ulatin
g. The observation ti
me is equ
al
to the arrival ti
me of the first peak after t
h
e
time-of-flight.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Ultra
s
oni
c To
m
ograph
y of Imm
e
rsion
Circula
r
Array b
y
Hype
rbol
a Algorithm
(Li
u
Yang)
67
2.2. Ultras
on
ic Propagati
on
Whe
n
the ultraso
n
ic is p
r
o
pagatin
g in the det
ection
-zone, the total pressu
re fiel
d in the
water i
s
cal
c
u
l
ated by solvi
ng the tran
sie
n
t pressu
re
wave equatio
n [15]:
2
22
11
t
td
m
p
p
qQ
ct
(1)
whe
r
e
the fluid de
nsity,
c
is
the so
und spe
ed,
t
p
is the total a
c
ou
stic pressu
re,
d
q
is
the
dipole sou
r
ce
,
m
Q
is the mono
pole
sou
r
ce. If there is
no
dipole
sou
r
ce
or mo
nopol
e
sou
r
ce, the
variable
s
d
q
and
m
Q
shoul
d be
0.
T
h
e
to
ta
l a
c
ou
s
t
ic
pr
es
su
e:
tb
p
pp
(2)
whe
r
e
p
stand
s for the prop
agation p
r
e
s
sure field,
b
p
sta
nds for th
e ba
ckgro
und p
r
e
s
sure field.
In the
circul
a
r
a
r
ray mod
e
l
, the
cylindri
c
al
wave
radi
ation i
s
a
dop
ted to
obtain
a
wide
rang
e
coveri
ng of the
ultraso
n
ic fiel
d a
nd mo
re
re
co
nstru
c
tion
dat
a. Then, the
cylindri
c
al
wa
ve
radiatio
n equ
ation is:
11
1
2
td
i
pp
np
q
Q
ct
r
(3)
whe
r
e
n
mean
s the no
rmal
vector of the radiation
sou
r
ce, and
11
1
2
ii
ii
pp
Qn
p
ct
r
(4)
whe
r
e
i
p
is the incid
ent pre
s
sure field whi
c
h is a fun
c
tio
n
of spa
c
e.
2.3. H
y
perbola Algorithm
A hyperb
o
la
may be
defin
ed eq
uivalen
t
ly as the l
o
cu
s of p
o
ints whe
r
e
the a
b
sol
u
te
value of the difference bet
wee
n
the dist
ances an
d
the two foc
i
is
a c
ons
ta
nt. Figure 2 sho
w
s the
geomet
ric
pa
ramete
rs of a
hyperbol
i
c
curve. In thi
s
figure,
1
F
and
2
F
are two fo
cal
p
o
ints of t
h
e
hyperb
o
la wit
h
the coo
r
din
a
tes of
,0
c
and
,0
c
.
O
is the origin
of t
he axes and the ce
nter
symmetri
c
po
int
for
th
e hyperb
o
lic curv
e.
1
V
an
d
2
V
with
the coo
r
din
a
tes
of
,0
a
and
,0
a
are two poi
nts of the inte
rse
c
tion of the
hori
z
ont
al
axis and th
e two bra
n
che
s
o
f
the hyperb
o
l
a,
respe
c
tively. Line
2
VQ
is perp
endi
cula
r to the ho
rizontal
axis and it
meets the
asymptotes at
point
Q
w
h
os
e c
o
or
d
i
n
a
t
es
a
r
e
,
ab
.
P
stand
s for any possible p
o
int o
n
each sid
e
of the
curv
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 1, March 2
015 : 65 – 75
68
Figure 2. Geometri
c par
a
m
eters of a hyperbol
a
Figure 3. Schematic di
ag
ram of hyperb
o
la algo
rithm
imaging
Acco
rdi
ng to the definition
of the hyperb
o
la:
12
2
PF
PF
a
(5)
whe
r
e
a
is the
distan
ce
bet
wee
n
the ve
rt
ex of a hyp
e
rbola a
nd th
e
origin
of the
axes,
whi
c
h i
s
obviou
s
ly a
consta
nt. Sup
posi
ng
1
F
an
d
2
F
can
be
repla
c
ed by t
w
o
re
ceiving tra
n
sd
uce
r
s,
then
P
ca
n be
con
s
i
dere
d
a
s
a certain
point
o
n
the te
sted
scattere
r
whi
c
h may reflect
the ultra
s
o
n
ic
pulse. The
r
ef
ore, it is n
o
t necessa
ry to care
ab
out the po
sition
of the tran
sm
itting transdu
cer
becau
se the
distan
ce fro
m
the transmitt
er to the
sc
atterer is
the same for eac
h
rec
e
iver [16].
Once the parameter
a
is o
b
tained, th
e
para
m
eter
b
can b
e
e
a
sily
indu
ced
from
the
equatio
n:
22
2
ab
c
(6)
whe
r
e
b
is the length of line
2
VQ
. Since para
m
eter
c
has al
so be
en obtaine
d from the
c
o
or
d
i
na
te
s
of
1
F
and
2
F
, the slope of the asymptotic line of the hyperb
o
la is
ba
.
So far, a d
e
t
ermine
d hyp
e
rbol
a
can
b
e
drawn thro
ugh the
follo
wing fu
nctio
n
(see
Figure 2):
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Ultra
s
oni
c To
m
ograph
y of Imm
e
rsion
Circula
r
Array b
y
Hype
rbol
a Algorithm
(Li
u
Yang)
69
22
22
1
0
,
b
>
0
xy
a
ab
(7)
The te
sted
scatterer is
su
p
posed o
n
the
hyperboli
c
curve, ho
weve
r only o
ne
hyperb
o
la
is in
sufficie
n
t
to locate
the
tested
scatterer. Since fan
-
sha
ped
ultra
s
onic
wave
s a
r
e reflecte
d b
y
the scatte
rer
in many
dire
ction
s
, hype
rbola
s
with dif
f
erent
param
eters a
nd
co
ordin
a
tes can
be
dra
w
n throug
h one tran
smi
tting transd
u
cer and
diffe
re
nt receivin
g transdu
cer p
a
irs.
Figure 3
sho
w
s the t
w
o
h
y
perboli
c
cu
rves inte
rse
c
ting at
P
an
d
'
P
. If the hype
rb
o
lic
curve
on
the
right i
s
exten
ded, o
ne
more inte
rse
c
tion
will
app
ear.
In fact, two h
y
perbol
as ma
y
have at
mo
st 4 p
o
ints of i
n
tersectio
n
a
nd the
o
reti
cal
l
y all of the
m
are li
kely to
be the
reflect
i
on
points.
With the chan
ging
of the tran
sm
itting
tran
sdu
c
er
in sequ
e
n
ce and
th
e perm
u
tation and
combi
nation
of the re
ceivi
ng tran
sd
uce
r
pairs,
a
sufficient nu
mbe
r
of the hyperbola is
obtain
ed
and supe
rimp
ose
d
togethe
r. Then, the image of
the tested scattere
r is re
con
s
truct
ed.
Hyperbola
ap
proa
ch
con
s
i
ders the
tran
sdu
c
e
r
s in g
r
oup
s of th
re
e, one
actin
g
as th
e
transmitter
a
nd two
as the
re
ceive
r
s.
T
he diffe
ren
c
e
s
b
e
twe
en th
e a
rrival tim
e
s of
the
scattered
sign
als at two
receive
r
s are
given by:
22
2
2
1
,
ij
p
p
p
i
p
i
p
j
p
j
tx
y
x
x
y
y
x
x
y
y
c
(8)
whe
r
e
,
ii
x
y
are th
e co
ordi
nate
s
of the first re
ceiver
and
,
j
j
x
y
the se
co
nd;
,
pp
x
y
stand fo
r the coo
r
din
a
tes o
f
the scatterer.
In the circula
r
array mod
e
l, the transd
u
cer
array ca
nn
ot be rotated
becau
se they
are all
embed
ded i
n
the wall
of the water
con
t
ainer. So, th
e numb
e
r
of received
sign
al pairs fo
r e
a
ch
time (or the h
y
perbol
as
ca
n be dra
w
n
)
i
s
given by:
1
LN
N
(9)
Whe
n
the
proce
s
s i
s
rep
eated fo
r e
a
c
h tran
smitting tra
n
sdu
c
e
r
, the n
u
mbe
r
of final
receiving si
gn
als shoul
d be
:
1
LN
N
N
(10)
The inten
s
ity of pixels at
the coo
r
din
a
tes
,
pp
x
y
in the recon
s
tru
c
ted i
m
age can be
cal
c
ulate
d
by
the
cross
co
rrelat
ion
s
of re
ceiving
si
gnal
pai
rs an
d the
su
mmation
o
peratio
n of
th
e
cro
s
s correlat
ions:
1
,
11
1
,,
NN
N
p
p
ni
nj
i
j
p
p
ni
j
k
in
j
n
Ix
y
R
t
x
y
(11)
whe
r
e
,
ni
nj
R
is the cro
s
s correlat
ion cal
c
ul
atio
n;
,
,
ni
nj
i
j
p
p
Rt
x
y
is the cross co
rrelation
value
of the receiving se
nsor pai
r
,
ij
3. Results a
nd Analy
s
is
3.1. Propaga
tion Simulation
Tran
sie
n
t si
mulation is
carrie
d out usin
g the COMSOL Mult
iphysi
cs®
so
ftware to
demon
strate the beh
aviour
of the ultra
s
o
n
ic wave
pro
pagatio
n insi
de the ci
rcula
r
array. Figu
re 4
sho
w
s the scattered a
c
ou
stic pre
s
sure d
i
stribut
io
n of 3 air bub
ble
s
at 6 different times of
5
1.
5
e
s
,
5
2
.
25e
s
,
5
3.
75
e
s
,
5
4.
5e
s
,
5
5.
25
e
s
,
5
8.
25
e
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 1, March 2
015 : 65 – 75
70
(a)
(b)
(b)
(d)
(e)
(f)
Figure 4. Tra
n
sie
n
t distrib
u
tion of the
sound p
r
e
s
sure scattered fo
r 3 air bu
bble
s
.
(a) a
nd (b
) sh
ow the tran
smitting and propag
ating
of the fan-sha
p
e
d
ultrasoni
c wave in the
water. (c) and
(d) represent
the wave refl
ection (mainl
y backwa
r
d
)
and diffra
c
tio
n
(mainly
forwa
r
d
)
due
to the blockin
g
of the air bu
bble
s
.
(e) a
n
d
(f) displ
a
y the se
con
dary reflection a
nd
multi-refle
c
tio
n
from the air bubble
s
. Du
ring the wh
ole
process, the
ultras
oni
c en
ergie
s
attenu
ate
contin
uou
sly
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Ultra
s
oni
c To
m
ograph
y of Imm
e
rsion
Circula
r
Array b
y
Hype
rbol
a Algorithm
(Li
u
Yang)
71
As can b
e
seen in thi
s
fi
gure,
3 teste
d
ro
und
air b
ubble
s
with the radiu
s
of
3mm a
r
e
longitudin
a
l a
rra
nge
d insid
e
the array and the
dista
n
c
e
s
between
the adja
c
ent
bubbl
es a
r
e
all
20mm. O
n
ly one tran
smitting tra
n
sdu
c
e
r
is visibl
e wh
ile the oth
e
r t
r
an
sdu
c
e
r
s a
r
e hi
dde
n whi
c
h
make
s th
e propag
ation p
r
oce
s
s mo
re
clear. Th
e
sca
l
i
ng facto
r
s of
the heig
h
t are not unifo
rm
for
the 6 differen
t
times, but it
is useful to get
a better ob
servatio
n. Th
e cou
p
ling m
edium in
side
th
e
circula
r
a
r
ray is the
water and the
radi
i of t
he air b
ubble
s
a
r
e al
l 3mm. For
constructin
g
a
n
infinite pro
p
a
gation
spa
c
e,
an impe
dan
ce mat
c
hin
g
layer is
set o
n
the oute
r
b
ound
ary of th
e
circula
r
a
r
ray. Con
s
e
que
ntly, the pro
p
a
gation
wave
pre
s
ent
s n
e
a
r
ly no
reflecti
on from
the o
u
ter
boun
dary.
3.2. Image Recons
truc
tio
n
The refle
c
tio
n
-mo
de sen
s
i
ng in UCT p
r
oce
s
se
s is b
a
se
d on the
time-of-flight
and the
amplitude me
asu
r
em
ent of reflection pat
h signal
s
due
to the existe
nce of an obj
ect betwe
en th
e
transmitting transdu
cer an
d
the
re
ceiving
tran
sd
uce
r
s.
Whe
n
the
tested a
r
ea
of th
e ci
rcula
r
a
r
ra
y
is blo
c
ked wi
th object
s
, the ultrasoni
c prop
agatio
n path is chan
ged an
d the amplitude of
the
received si
gn
al is de
cre
a
sed.
Figure 5
sh
o
w
s
3 te
sted
air b
ubble
s
a
nd the
re
con
s
tru
c
ted im
a
ges. T
he p
r
e
s
sure
of
each
b
ubble
is set
to be equal
to
the atmosp
he
ric pre
s
sure
5
1.
01
e
P
a
sin
c
e th
e hyd
r
a
u
lic
pre
s
sure is
n
egligible
he
re
. And the de
n
s
ity of the bu
bble i
s
3
1.195
k
g
m
(satu
r
ation state) and
the sou
nd sp
eed in the air
bubbl
e is
344
m
s
As se
en in Fi
gure
5, the re
con
s
tru
c
tion i
n
tensit
y of the rou
nd bu
bb
le at the ce
nter of th
e
array is
high
e
r
than
the oth
e
r two b
ubble
s
. So, it
is
co
nclu
ded th
at the
sen
s
itivity of the ci
rcula
r
array increa
ses with its
rad
i
us an
d the p
eak lo
cate
s a
t
the centre .
Figure 6
di
splays
5
rou
n
d
ai
r b
ubbl
e
s
a
nd th
eir reco
nstructe
d
imag
es.
Th
e teste
d
bubbl
es
with a radi
us of 5
mm are
spa
c
ed aroun
d the
cente
r
of th
e circul
ar a
r
ray. The hori
z
ontal
and lon
g
itudi
nal dista
n
ces of each ai
r
bubbl
e are
al
l 20mm. Wit
h
the increa
se of the test
ed
bubbl
e num
b
e
rs, the
outlin
e of the sin
g
le bubbl
e ima
ge is di
storte
d whe
r
e
a
s its size
dimen
s
i
o
n
sho
w
s better
agre
e
me
nt wi
th the tested bubbl
e.
In the above experim
ent, the ultra
s
oni
c
wavele
ngth i
n
the wate
r is approxim
atel
y 3mm.
By comp
arin
g Figu
re
5
wit
h
Figu
re
6, it
is fou
nd th
at
whe
n
the
ultraso
n
ic wavel
ength i
s
clo
s
e to
the dim
e
n
s
io
ns
of the
test
ed ai
r
bub
ble
s
, the
si
ze
of
the
re
con
s
tructed
bu
bble
imag
es do
n
o
t
agre
e
with th
e bubbl
e si
ze
very well. Howeve
r, the outline sho
w
s a good ag
re
ement with e
a
ch
other in th
e two figu
re
s. Whe
n
the ult
r
asoni
c
wave
length is
sm
a
ller than th
e dimen
s
ion
of the
tested air b
u
b
b
les, the correct si
ze dime
nsio
n ca
n be
recon
s
tru
c
ted
.
In the mean
time, much
smalle
r ai
r b
ubble
s
can
be dete
c
ted
while i
n
crea
sing th
e
freque
ncy of the ultrasoni
c wave. Figure
7 shows 9 ai
r bubbl
es wit
h
different si
zes and
sha
p
e
s
in a ci
rcula
r
region
with a
radiu
s
of
5m
m. T
he radii
of the ro
und
bubbl
es i
n
th
e middle
col
u
mn
are 50
0
m
and the other 2 smaller round
bubbl
es in the 1st and 3rd
column
s are 300
m
. The
major an
d mi
nor axial l
eng
ths of
the
re
maining
4
elli
pse
s
are 5
0
0
m
and
30
0
m
re
spectively.
The ellip
se
s are
spa
c
ed i
n
tilted 45 degre
e
angl
e.
In order to i
dentify the air bubbl
es at
the
hund
red
-
mi
cron scal
e, 5
M
Hz freq
uen
cy tran
sd
uce
r
s
are
empl
o
y
ed as the
element
s of
the
circula
r
array.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 1, March 2
015 : 65 – 75
72
(a)
(b)
(c)
(d)
Figure 5. 3 air bubbl
es a
n
d
the reconst
r
u
c
t
ed imag
es
obtaine
d by hyperbol
a algo
rithm and
threshold filte
r
ing (Unit: mm).
(a). 3 teste
d
round ai
r bub
b
l
es are longit
udinal a
r
rang
ed within the i
mmersion ult
r
aso
n
ic a
r
ray,
the radiu
s
of each bub
ble i
s
3mm an
d the dist
an
ce
s b
e
twee
n adja
c
ent bubbl
es a
r
e 20mm. (b).
The co
rrespo
nding recon
s
tructe
d imag
e of the 3 air
bu
bble
s
. (c) a
n
d
(d). Imag
es after threshol
d
filtering
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Ultra
s
oni
c To
m
ograph
y of Imm
e
rsion
Circula
r
Array b
y
Hype
rbol
a Algorithm
(Li
u
Yang)
73
(a)
(b)
(c)
(d)
Figure 6. Five (5)
roun
d ai
r bubbl
es a
n
d
the corre
s
po
nding recon
s
tructe
d imag
e
s
obtain
ed by
hyperb
o
la alg
o
rithm (Unit: mm).
(a). 5 tested
bubbl
es
with the radi
us of 5
mm are
spa
c
ed aro
und th
e cente
r
of the circula
r
arra
y.
The ho
rizonta
l
distan
ce an
d longitudin
a
l
distan
ce
of e
a
ch a
d
ja
cent
air bub
ble a
r
e
20mm. (b).
The co
rrespo
nding recon
s
tructe
d imag
e of the 5 air bu
bble
s
. (c). Th
e recon
s
tru
c
t
ed image afte
r
threshold filte
r
ing. (d
). Gra
y
scal
e
image
with better
co
ntrast
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 1, March 2
015 : 65 – 75
74
(a)
(b)
(c
)
(d)
Figure 7. 9 air bubbl
es
with different si
zes an
d the co
rre
sp
ondi
ng reco
nstructe
d image
s
obtaine
d by hyperbol
a algo
rithm (Unit: mm).
(a). 9 teste
d
air bub
ble
s
wi
th different si
ze
s and
sha
p
e
s. (b). Th
e corre
s
p
ondin
g
recon
s
tru
c
ted
image of 9 air bubbl
es. (c)
and (d
). Imag
es after threshold filtering
Hyperbola m
e
thod emp
h
a
s
izes t
he cro
s
s-correlatio
n
calculat
ion u
nder the
circu
m
stan
ce
of the permu
tation and th
e com
b
inatio
n of all pos
si
ble re
ceiving
signal
s. The
most impo
rtant
benefit is th
at the numb
e
r of pote
n
tial com
b
inati
ons
ha
s gon
e up to
12
NN
N
. The
comp
aratively clea
r
outlin
e of th
e recon
s
tru
c
ted
im
ag
e can
be
obta
i
ned
by the
th
reshold
filteri
ng.
Ho
wever, th
e mo
st effe
ctive app
ro
a
c
h fo
r im
p
r
o
v
ing
the re
solution
a
nd contrast of
the
recon
s
tru
c
ted
images i
s
increa
sing the freque
ncy an
d numbe
r of the transdu
ce
r.
4. Conclusio
n
A non
-inva
s
ive UCT
syste
m
for
extracti
ng the
cro
s
s-se
ctional
ima
ges of the
two-ph
ase
flow ba
sed o
n
circula
r
a
r
ray is develop
ed and in
ve
st
igated in this
study. Two
-
p
hase flow im
age
s
with different
gas di
strib
u
tions a
r
e reco
nstru
c
t
ed
usi
ng the hype
rbola alg
o
rith
m and their d
i
stinct
bubbl
e dia
m
eters an
d di
spe
r
sed
pha
se
distri
butio
n are id
entified. Tran
sient
simul
a
tionof
the
ultrasoni
c p
r
o
pagatio
n is
condu
cted
by usin
g the
CO
MSOL Multip
hysics®
soft
ware to illu
strate
the multiple reflection
path
s
of the ultra
s
oni
c
waves
and obt
ain th
e inspectio
n
sign
als of e
a
c
h
element. Refection m
ode
is utilized to
extract
s
in
formation of the
con
c
e
n
tratio
n profile
of th
e
Evaluation Warning : The document was created with Spire.PDF for Python.