TELKOM
NIKA
, Vol.11, No
.1, Janua
ry 2013, pp. 73
~82
ISSN: 2302-4
046
73
Re
cei
v
ed Au
gust 29, 20
12
; Revi
sed
No
vem
ber 2
2
, 2012; Accepte
d
No
vem
ber
28, 2012
Measur
e
d Data Processing Method for Relative Motions
Between Two Side-by-side Ships
Ping-an Shi*
1,2
, Jia-
w
e
i Y
e
1
1
School of Civ
il
Engin
eeri
ng a
nd T
r
ansportati
on, Sout
h C
h
in
a Univ
ersit
y
of
T
e
chnolog
y, Guan
gzh
ou,
Chin
a,
2
No.1 Divisi
on,
Naval Arms C
o
mmandi
ng Aca
dem
y
,
Gu
angz
hou, Ch
in
a
*corres
pon
di
ng
author, e-mai
l
: pashi
@12
6
.co
m
A
b
st
r
a
ct
In order to design and implem
en
t a wave compensation
system
to
reduce the relativ
e
motion
betw
een tw
o side-by-si
de sh
ip
s in w
a
ves, a new
metho
d
to process
me
asu
r
ed data of shi
p
mo
de
l test w
i
t
h
contact
me
asu
r
ement to
stud
y the
c
haracte
ristics of re
lati
ve
moti
on w
a
s
pres
ented. T
h
e refer
ence
co
-
ordinate syst
em
s
and relativ
e
m
o
tions wer
e
defined, and the sc
he
m
e
of the
m
o
del test was described.
T
hen the E
m
p
i
rical M
ode
D
e
co
mp
ositio
n (
E
MD) ada
ptiv
e filter w
e
re d
e
sig
ned, the fr
equ
ency d
o
ma
in
in
te
g
r
a
t
io
n tra
n
sfo
r
m
me
th
o
d
b
a
s
ed
up
on
Fast Fo
u
r
ie
r Transfo
rm
(FFT) we
re
e
s
tab
l
ish
e
d
.
Th
e
p
r
o
c
edu
re
to transform a
cceler
a
tion si
g
nal int
o
disp
lac
e
ment
w
a
s propose
d
and ver
i
fied, and the p
r
ocessi
ng resu
l
t
s
w
i
th and w
i
th
o
u
t EMD a
d
a
p
ti
ve filter w
e
re
c
o
mpar
ed. F
i
n
a
l
l
y, the re
lative
moti
ons
c
ons
is
tent w
i
th real
ity
w
e
re acquir
ed,
w
h
ich indic
a
te
s this meth
od i
s
effective for me
asur
ed d
a
ta
processi
ng.
Key
w
ords
: relative m
o
tion, ship m
odel test, acce
leratio
n
, fre
quen
cy dom
ain integratio
n,
em
pirical m
ode de
com
position (EMD)
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
Whe
n
two ve
ssels
are
po
sitioned
side
-b
y-side
at sea
to do repleni
shme
nt, the relative
motions b
e
tween th
em
are of
co
nsi
derable i
m
po
rt
a
n
ce.
Du
e to
the
clo
s
e
pro
x
imity of the
two
vessels, hydrodynami
c
intera
cti
on bet
ween them is
highly increa
sed a
nd it ha
s great effect
on
relative motio
n
. The larg
e relative motio
n
may cau
s
e
repleni
sh
me
nt work difficulties, the ca
rgo
whi
c
h i
s
de
scendin
g
will
be
collidi
ng
with
the sh
ip b
oard whi
c
h i
s
ascen
d
ing
and t
he cargo
whi
c
h
has lo
ade
d o
n
ship b
oard
will be su
sp
ende
d in the
air again,
so
metimes eve
n
leadin
g
to the
colli
sion bet
wee
n
the su
perstru
ct
ures of two ve
ssel
s. Becau
s
e of the
s
e
seriou
s
pro
b
le
ms
cau
s
e
d
by th
e hyd
r
odyna
mic i
n
tera
ctio
n effect, it
is
very important to
s
t
udy the relative motion
behavio
rs be
tween
two
side-by
-si
de
moored ve
ssels[1,2]. Th
e
hydrodyna
mic i
n
tera
cti
ons
betwe
en m
u
ltiple b
odie
s
h
a
ve be
en
rep
o
rted
by
ma
n
y
repo
rte
r
s[3,
4], but only
several
autho
rs
dealt with the
corre
s
po
ndi
ng pro
b
lem
s
for the shi
p
-li
k
e bo
die
s
an
d few autho
rs discu
s
sed t
h
e
relative motio
n
s for vari
ou
s loading p
o
siti
on.
In orde
r to un
derstand th
e relative motio
n
cha
r
a
c
teri
st
ics
of two si
d
e
-by-sid
e
po
sitioned
ship
s in wav
e
s, the co
ntact mea
s
u
r
e
m
ent method
is used in
ship mo
del e
x
perime
n
t under
different
wav
e
condition
s.
In the exp
e
r
iment, the
a
c
celeromete
rs
were u
s
e
d
to mea
s
u
r
e
th
e
accele
ration
s of surg
e, sway and heav
e of each
sh
i
p
model, so it is nece
s
sa
ry to convert the
accele
ration
sign
al into displa
cem
ent. It needs do
ubl
e integratio
n to transfo
rm accele
ration i
n
to
displ
a
cement
. But becau
se of the n
o
ise
s
em
b
e
d
ded in the
measured d
a
ta, the resulting
displ
a
cement
by di
re
ct d
o
uble i
n
tegration
i
s
hardly useful. The
me
thod
to filter th
e n
o
ises and
improve the
accura
cy of resultin
g di
spla
cem
ent is nee
ded. T
o
obtain the
relative motion
betwe
en two
side
-by-sid
e
ship
s, the ref
e
ren
c
e
co
ordi
nate sy
stem
s and relative
motions
sh
ould
be d
e
fined.
The p
ape
r i
s
dedi
cate
d t
o
re
solve
th
ese
proble
m
s, an
d to
prese
n
t an
effective
method to p
r
ocess m
e
a
s
ured
data of
model te
st
to acq
u
ire th
e ch
ara
c
te
ristics of relative
motions b
e
tween two
side
-by-sid
e
shi
p
s
in wave
s.
This
pap
er i
s
o
r
ga
nized
as follows.
The
next section
de
scri
bes the d
e
finition of
referen
c
e co
-ordin
ate syst
em and rel
a
tive motion an
d the sch
em
e of model test. The se
ct
ion
that follows d
e
scrib
e
s EM
D Adaptive Fi
lter and
Freq
uen
cy Domai
n
Integration.
Then the met
hod
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 73 – 82
74
to process m
eassu
red
dat
a to
ca
l
c
ulate
the relative motion betwe
en
two side
-by-sid
e
ship
s
i
s
descri
bed a
n
d
the relative motion is a
c
q
u
ired.
In the last se
ction, concl
u
si
on
s are made.
2. Definition
of Rela
tiv
e
Motion and
Model Tes
t
Scheme
Two
side
-by-side
po
sition
ed vessel
s
show
different
cha
r
a
c
teri
sti
cs f
r
om o
n
e
singl
e
vessel in
wa
ves. In
ord
e
r to
study the
ch
arac
te
risti
c
s of
relative
motion
bet
ween t
w
o
sid
e
-
by-
side
po
sition
ed ship
s in
waves, it is
essential
to defin
e the referen
c
e
coo
r
di
nate
system
and
the
relative motio
n
. The relative motion
ch
a
r
acte
ri
st
ics were
studie
d
b
y
model te
st, so the
sch
e
m
e
of ship mod
e
l
test is also b
r
iefly descri
b
ed here.
2.1. Refe
ren
ce Co
-ordina
t
e Sy
stem
In orde
r to de
scribe the
rel
a
tive motion betwe
en two
side
-be
-
side
ship
s in wave
s, three
sets
of right
-h
ande
d orth
og
onal
coo
r
din
a
t
e system
s a
r
e ado
pted, which
are
sh
o
w
n a
s
Fig.1.
The
O-XYZ sy
ste
m
is the ine
r
tial coo
r
dinat
e system. T
he O
A
-X
A
-Y
A
-Z
A
and O
B
-X
B
-Y
B
-Z
B
are t
h
e
oscillatory co
ordin
a
te syst
ems fixed wit
h
re
spe
c
t to ship A and
ship B, resp
ectively. The O-XY
plane re
sts on
the cal
m
wate
r surf
a
c
e,
the
X-axi
s
p
o
ints f
o
rward
and
the
Z-axis vertically
upward. T
he
oscillatory
coordinate sy
stems
O
A
-X
A
Y
A
Z
A
and O
B
-X
B
Y
B
Z
B
are u
s
e
d
to de
scrib
e
the
ship’
s
motion
in six degree
s of freed
om.
2.2. Relativ
e
Motion Bet
w
een T
w
o
Sid
e
-By
-
Side Ships
Each vessel t
a
ke
s 6-DO
F(degree of fre
edom
)
motio
n
s. Assumin
g
that the loading point
is on Ship
-B. The po
sition
of the loading point is
shown as Fig
u
re 1. The
coordi
nate
s
of the
positio
n
with respe
c
t
to
ea
ch ship’
s
refe
ren
c
e coo
r
din
a
tes system are (x
a
, y
a
, z
a
) and
(x
b
, y
b
, z
b
),
r
e
spec
tively.
Figure 1. The gene
ral coo
r
dinate
s
of the loadin
g
poi
nt
Whe
n
Ship-B
takes 6
-
DOF
motions, the positio
n vecto
r
of the loadin
g
point in O
B
-X
B
-Y
B
-
Z
B
coordinate
system is a
s
follows:
(
b
1
+
b
b
z
5
+
b
b
y
6
,
b
2
+
b
b
x
6
-
b
b
z
4
,
b
3
+
b
b
y
4
-
b
b
x
5
)
Whe
n
Ship-A
takes 6
-
DOF
motions, the positio
n vecto
r
of the loadin
g
point in O
A
-X
A
-Y
A
-
Z
A
coordinate
system is a
s
follows:
(
a
1
+
a
a
z
5
+
a
a
y
6
,
a
2
+
a
a
x
6
-
a
a
z
4
,
a
3
+
a
a
y
4
-
a
a
x
5
)
Thus,
the l
o
n
g
itudinal, l
a
te
ral a
n
d
vertical re
l
a
tive m
o
tion (re
p
re
sented
by X,
Y and
Z)
betwe
en ship
A and ship B
at any position ca
n be
cal
c
ulate
d
by the followin
g
three eq
uation
s
:
)
(
)
(
6
5
1
6
5
1
a
a
a
a
a
b
b
b
b
b
y
z
y
z
X
(1)
Evaluation Warning : The document was created with Spire.PDF for Python.
TEL
K
deno
2.3.
M
by-si
d
the
C
dime
with
wav
e
the r
e
simil
a
posit
i
give
n
Figur
e
incid
e
the h
heav
e
moti
o
trans
3. E
M
the d
twic
e
integ
r
bias
con
d
frequ
meth
upon
A
K
OM
NIKA
(
2
b
Y
(
3
b
Z
Whe
r
e, t
h
t
e
sur
ge,
sw
M
odel Test
S
The obje
c
d
e mo
ored
s
C
oa
stal a
nd
O
n
s
ion of the
a
wave
gen
e
s; the ot
her
e
flec
ted wa
v
a
rity criteria,
i
o
ned
sh
ip
m
n
in Tabl
e 1.
e
2. Two shi
p
s
The m
o
d
e
nt wave
an
g
isto
ry of the
e
motions o
f
o
ns we
re
m
e
form accele
r
M
D Adap
tiv
e
The su
rg
esired relati
v
e
to
be trans
f
r
atio
n is not
an
d misali
g
itioning
can
e
n
cy of the
od ba
se
d u
p
FF
T used t
o
A
cceler
o
met
e
Meas
u
6
4
b
b
b
b
x
z
5
4
b
b
b
b
x
y
h
e subscri
p
t
ay, heave,
r
o
S
chem
e
c
tive of the
m
s
hip
s
in di
ffe
r
O
ffshore En
g
tank is
32
m
×
er
a
t
or
a
t
o
n
end
is
pa
v
e
v
e. Th
e two
t
he simi
lari
t
m
o
dels in
ex
p
mo
del
s po
s
ide in
wave
d
el t
e
st
wa
s
g
les
,
differe
n
wave el
eva
t
f
the two shi
e
asure
d
by ti
l
r
atio
n into di
s
e
F
ilter and
F
e, sway
an
d
v
e motion
of
f
or
med
into
too goo
d a
n
g
nment of
be reali
z
ed,
filter i
s
not
p
on FFT
a
n
o
tran
sform f
e
r
I
S
u
red Da
ta Pr
o
(
)
4
2
a
a
b
(
)
4
3
a
a
b
y
t
of
a
and
b
o
ll, pitch an
d
m
o
del test
i
s
r
ent wave
c
o
g
inee
ring
La
b
×
18
m×1m a
n
n
e end, whi
c
e
d with the
1
shi
p
mo
del
s
t
y
scal
e
is
2
perim
ent. T
h
s
itione
d sid
e
s
con
ducte
d
n
t wave freq
u
t
ion wa
s m
e
a
ps
w
e
r
e
me
a
l
t senso
r
s.
S
s
pl
acement
.
F
requen
c
y
D
he
ave
mo
ti
the lo
adin
g
displ
a
cem
e
n
n
d even h
a
r
the accel
e
r
as prese
n
t
e
eas
y
. In or
d
n
d EMD is
p
iltered accel
e
Wave Sen
s
T
ilt Sensor
S
SN: 2302-4
0
o
ce
s
s
ing
Me
)
6
a
a
a
a
x
z
)
5
a
a
a
x
y
den
ote shi
p
d
yaw
,
r
e
s
p
e
c
to
stud
y the
o
nditions. Th
b
or
a
t
or
y in
S
n
d the
dept
h
c
h ca
n
p
r
od
u
:7 wave dis
s
s
u
s
e
d
i
n
t
h
e
2
5. Figure 2
h
e m
a
in
pa
r
a
e
-by
-
Tabl
e
Ite
m
L
p
B
LC
G
M
Z
g
Displa
c
un
de
r diff
e
u
en
cie
s
and
a
su
red by
a
w
a
sure
d by a
c
S
o, in orde
r t
o
D
omain Int
e
on
s a
r
e me
a
po
sition, the
n
t, but the r
e
r
dly u
s
eful.
T
r
om
eters,
a
e
d by Go
dh
a
v
d
er
to
imp
r
o
v
p
re
sente
d
.,
a
e
ration into
d
s
or
0
46
thod Fo
r Re
l
p
A a
nd s
h
i
p
c
tively.
relative mo
t
e mod
e
l tes
t
S
o
u
th
C
h
ina
of the wate
r
u
ce t
w
o-dim
e
s
ipatio
n ram
p
e
experim
e
n
t
shows the
a
meters
for
t
e
1. Main P
a
m
unit
s
p
p
m
B
m
1
G m
-
M
m
g
m
ement
m
3
3
e
rent wa
v
e
c
different w
a
w
ave h
e
ight
c
celeromete
o
get the rel
a
e
gration
a
su
red
by
t
h
accele
ratio
n
e
sulting di
sp
T
o red
u
ce t
y
a
nd ensure
v
en [5]. But
v
e the a
c
cu
r
a
nd frequ
en
c
d
ispl
acemen
l
ati
v
e Moti
o
n
p
B, a
nd
th
e
t
ion ch
ara
c
t
e
t
wa
s c
a
r
r
ie
d
Univ
e
r
sity
o
f
r
i
s
0.8m. T
h
e
e
ns
io
na
l re
g
p
, whic
h
c
a
n
t
wer
e
ma
d
e
p
hoto of th
e
t
he two
shi
p
a
rameters o
f
Suppl
y
i
ng
s
hip model
72 2.88
1
3.9 0.56
1.25
-0.05
3.1 0.16
4.5 0.19
3
900
0.249
c
on
dition
s,
i
a
ve height
s.
rec
o
rder; th
e
rs, and the
r
a
tive motion
,
r
e
e ac
cele
r
o
n
sign
als
ha
v
l
a
cem
ent g
o
y
pic
a
l er
rors
a stable i
t
he sel
e
ct
io
n
r
acy, a
ne
w
c
y domanin
t.
n
s …
(Ping-
a
e
subs
cr
ipt
o
e
ri
st
ic
s of
t
w
o
d
out in the t
a
f
Technol
og
y
e
t
ank i
s
eq
u
g
ul
a
r
an
d i
r
r
e
n
eliminate 9
e
ac
co
r
d
in
g
e
two sid
e
-
b
p
s and mod
e
f
Ship Model
Receiving
ship mo
d
77 3.
0
15.7 0.
6
-1.64
-0.
0
4.46 0.
1
4.37 0.
1
2673
0.
1
i
ncludi
ng di
f
In the exper
e
surge,
sw
a
r
ol
l, pitch an
,
it is neces
s
o
meters
. To
o
v
e to be inte
g
o
t by
direc
t
d
lik
e
se
ns
or
nt
egratio
n
,
n
of the thr
e
i
n
tegral
tra
n
int
e
gratio
n
b
a
n Sh
i
)
75
(2)
(3)
o
f 1~
6
o
sid
e
-
a
nk
o
f
y
. The
u
ippe
d
e
gul
ar
0% of
to t
h
e
y
-si
de
e
ls a
r
e
s
d
el
0
8
6
3
0
7
1
8
1
8
1
7
f
ferent
i
men
t,
a
y a
nd
d yaw
s
ar
y to
o
btain
g
rate
d
d
ou
ble
no
ise,
sig
n
al
e
sh
old
n
sf
orm
b
ase
d
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 73 – 82
76
3.1. Empiric
a
l Mode De
c
o
mpos
ition(EMD) Adaptiv
e
Filter
EMD is used
to filter the
noises
embe
dded
i
n
the a
c
celeration
si
gnal a
daptive
l
y. EMD
has b
een
wi
dely use
d
to analyze non
-stationa
ry an
d nonlin
ear
signal by de
compo
s
ing int
o
a
seri
es of intri
n
si
c mo
de fu
nction
s
(IMF
s)
and
a tre
n
d
functio
n
[7]. EMD i
s
ad
a
p
tive, and IM
Fs
become the
basi
s
rep
r
esenting the u
nderlyin
g dat
a.
An IMF is a function t
hat sati
sfies
two
con
d
ition
s
: (1
) in the whol
e data set, the numbe
r
of extrema an
d the numbe
r o
f
zero
cro
s
sin
g
s
must eithe
r
e
qual o
r
differ
at most by on
e;
and (2
) at
any point, the
mean value
of the envelo
p
e
defined by th
e local maxim
a
and the env
elope d
e
fined
by the local minima is
zero.[8,9]
For a
n
y give
n sig
nal
)
(
t
x
, EMD i
s
impl
e
m
ented th
ro
ugh a
sifting
pro
c
e
s
s tha
t
is
summ
ari
z
ed as
follo
ws:[10
]
(1)
To id
entify all the lo
cal
extrema. Se
perately
co
nn
ect all th
e ma
xima and
min
i
ma with
natural
cubi
c
splin
e lines to
form the upp
er and lo
we
r
envelop
es.
(2)
Find
the mean of
the envelop
es as
)
(
1
t
m
, and ta
ke th
e differe
nce
betwe
en the
data
and the mea
n
as
)
(
1
t
h
, which is
the proto
-
IMF
:
)
(
)
(
)
(
1
1
t
m
t
x
t
h
(4)
(3) Ch
eck th
e
proto
-
IMF
)
(
1
t
h
a
gain
s
t the d
e
finition of IMF
and the
sto
p
p
age
crite
r
ion
to
determi
ne if it is an IMF. If
)
(
1
t
h
doe
s not sati
sfy the definition, repe
at ste
p
1 to 3 on
)
(
1
t
h
:
)
(
)
(
)
(
11
1
11
t
m
t
h
t
h
(5)
...
...
Rep
eat step
1 to 3 on
)
(
1
t
h
i
till i
t
satisfies the
definit
ion after k times of si
fting.
)
(
)
(
)
(
1
)
1
(
1
1
t
m
t
h
t
h
k
k
k
(6)
If
k
h
1
doe
s satisf
y the definition, assi
gn it as an IMF comp
onent,
)
(
1
t
c
.
k
h
t
c
1
1
)
(
(7)
The first resi
d
ue ca
n also b
e
got as follo
ws:
)
(
)
(
)
(
1
1
t
c
t
x
t
r
(8)
Now, the firs
t
IMF
c
o
mponent(
)
(
1
t
c
) ha
s be
e
n
got
from
th
e o
r
iginal
sig
nal, it
contai
n
s
the best si
gn
al scale or th
e sho
r
test pe
riod com
pon
e
n
t. The stopp
age criteri
on i
s
as follo
ws:
SD
h
t
h
t
h
t
k
k
k
2
)
1
(
1
2
1
)
1
(
1
)]
(
)
(
[
(9)
Whe
r
e,
)
(
1
t
h
k
is the result of x(t) after k times
of sifting, SD rang
e betwee
n
0.2 and 0.3.
(4) Repe
at th
e op
eratio
n from
step
1 to
3 on
the
re
sid
ue, r(t)
= x
(
t)
– c(t), a
s
the
origin
al
sign
al.
The
operation en
ds wh
en
th
e
re
sidu
e,
)
(
t
r
n
, becom
es a
m
onotoni
c fu
n
c
tion
or a
function
cont
aining o
n
ly one intern
al extremum
fro
m
whi
c
h no mo
re IMF can be
extracted.
So, any given signal x(t)
ca
n be de
comp
ose
d
by EMD as follows:
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Measured Da
ta Proce
s
sing
Method Fo
r Relati
ve Moti
ons …
(Ping-an Shi)
77
n
i
i
n
t
c
t
r
t
x
1
)
(
)
(
)
(
(10
)
whe
r
e
)
(
t
r
n
stand
s for a resi
du
al “tren
d
” a
n
d
the “mode
s”
}
,...,
1
),
(
{
n
i
t
c
i
are con
s
trai
ned to
be ze
ro
-mea
n amplitude
modulatio
n freque
ncy mod
u
lation waveform
s.
It can b
e
see
n
from th
e E
M
D, ea
ch
IMF ha
s differe
nt ch
ara
c
te
ristic of fre
que
n
c
y ba
nd,
and th
e
re
sid
ue i
s
a m
ean
“tre
nd
” o
r
co
nstant. F
r
e
q
u
ency
ban
d of
IMF d
e
crea
ses f
r
om
c
1
to
c
n
,
with
c
1
th
e hi
ghe
st and
c
n
the lowest.In the decom
positio
n, the
IMF function depend on t
he
sign
al itself, different
sign
al can
gen
erate di
ffere
nt
IMF functio
n
, so
the EM
D i
s
a
daptiv
e.
Acco
rdi
ng
to
this
featu
r
e, an
EM
D ba
sed self-ada
ptive filter
can
be im
pleme
n
ted. Th
e id
ea
of
the adaptive filter is to decompo
se the signal by EMD, and then
the IMF compon
ents are filtered
according to the ban
d ch
aracte
ri
stics of the noi
se.[11]
High-Pa
ss F
ilter
. If the noise i
n
the
si
gnal i
s
lo
w-freque
ncy o
s
ci
llation noi
se,
then a
high-pa
ss filter
ba
sed
up
on EM
D
de
compo
s
ition
can b
e
con
s
tructed
to filter the
noi
se. T
h
e
result of the filtering can be
expresse
d a
s
follows:
1
)
(
)
(
i
i
hp
t
c
t
x
(11
)
Lo
w
-
Pa
ss Filter
. If the n
o
ise
in the
si
gnal i
s
hi
gh-f
r
equ
en
cy o
s
cillation noi
se,
then
a
low-pa
ss filter base
d
upo
n EMD de
comp
osition
can b
e
con
s
tru
c
ted
to filter the noise. The
re
sult
of the filtering can be exp
r
e
s
sed a
s
follo
ws:
)
(
)
(
)
(
t
r
t
c
t
x
n
K
i
i
lp
(12
)
Band
-Pas
s Filter
. If the noise i
n
the
signal
co
nta
i
ns both
hig
h
and lo
w freque
ncy
oscillation n
o
i
s
e, then
a ba
nd-p
a
ss filter
based
u
pon
EMD de
com
p
osition
ca
n b
e
co
nstructe
d
to
filter the noise. The re
sult of the
filtering can be exp
r
e
s
sed a
s
follo
ws:
i
i
bp
t
c
t
x
)
(
)
(
(13
)
3.2. Frequenc
y
Domain Integration
The p
r
in
ciple
of freque
ncy
domain i
n
te
gration i
s
a
s
follows: The
FFT is a
pplie
d to the
measured time domain
accele
ration
seq
uen
ce, a
nd then the
resultin
g freque
ncy do
main
accele
ration
seq
uen
ce i
s
integrate
d
twi
c
e to get disp
lacem
ent.
Con
s
id
er a time domai
n si
gnal x(t) of time
wind
ows l
ength T, whi
c
h is sam
p
led
N times
to obtain the discrete time
seri
es x(n). T
hen
the no
rm
alize
d
DFT of
x(n) is defin
e
d
as:
1
0
)
2
(
)
(
)]
(
[
)
(
N
n
nk
N
j
e
n
x
n
x
DFT
k
X
(14
)
Whe
r
e, n and
k take value
s
of 0,1,2,……,(N-1).
DFT is im
ple
m
ented
with Fast Fou
r
ie
r Tran
sfo
r
m(F
F
T
) algo
rithm.
X(k) i
s
a co
mplex
valued serie
s
of length N(freque
ncy spe
c
trum
):
)]
,
(
),...,
,
(
),
,
[(
)]
(
[
)
(
1
1
2
2
0
0
N
N
jb
a
jb
a
jb
a
n
x
DFT
k
X
(15
)
The amplitud
e, circular fre
quen
cy and i
n
it
ial phase a
ngle of each harm
oni
c co
mpone
nt
of x(n) ca
n be
calculated a
s
follows:
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 73 – 82
78
k
k
k
k
k
k
k
a
b
T
k
w
b
a
A
arctan
/
2
2
2
(16
)
Acco
rdi
ng to
the si
gnal
superpo
sition
prin
cipl
e, a
n
y peri
odi
c si
g
nal can b
e
o
b
tained
throug
h the superpo
sition
of certai
n ha
rmonic
sig
nal.
If the accel
e
ration
sign
al is expresse
d as
formula (17
)
, the displ
a
ce
ment sign
al can be expressed a
s
form
ul
a (18
)
.
)
cos(
......
)
cos(
)
cos(
1
1
1
1
0
0
1
1
0
N
N
a
N
a
a
a
a
a
t
w
A
t
w
A
t
w
A
a
(17
)
)
cos(
......
)
cos(
)
cos(
1
1
1
1
0
0
1
1
0
N
N
d
N
d
d
d
d
d
t
w
A
t
w
A
t
w
A
d
(18
)
In the freq
ue
ncy dom
ain, t
he di
spla
cem
ent ca
n be
o
b
tained
by scaling the
acceleratio
n
by the squa
re of the
freque
ncy. Bo
th t
he ampli
t
ude an
d p
hase relatio
n
shi
p
s
between
accele
ration
and di
spla
ce
ment are
sho
w
n a
s
formul
a (19
)
and (2
0):
2
/
i
a
d
w
A
A
i
i
(19
)
i
i
a
d
(20
)
In orde
r to att
enuate th
e effect of ran
d
o
m
noises
em
bedd
ed in a
c
cele
ration
sig
nal, it is
necessa
ry to filter them. In
the frequ
e
n
cy dom
ain,
it can be
do
ne by setting
the potion
of
freque
ncy t
h
a
t
need
to b
e
f
iltered to
zero
. Freq
uen
cy
domain
filteri
ng m
e
thod
ca
n be
exp
r
e
s
sed
as
follows
:
1
0
/
2
2
)
(
)
(
)
2
(
1
)
(
N
k
N
nk
j
e
k
X
k
H
f
n
y
(21
)
others
f
f
k
f
k
H
u
d
,
0
)
(
,
1
)
(
(22
)
In whi
c
h, k, n
and
r takes v
a
lu
e
s
of 0, 1,
2, …, N-1;
f
d
and
f
u
are the
lowe
r
cut-off
freque
ncy a
n
d
uppe
r
cut-off freque
ncy
of
band
-pa
s
s filter; X(k) is
the
FFT tran
sformation of tim
e
dom
ain
sig
nal
x(
n)
;
△
f is fre
quen
cy re
sol
u
tion; H(k) i
s
the band
pa
ss filter frequen
cy re
spon
se f
unctio
n
.
Finally,
the resultin
g
fre
q
u
ency domai
n integratio
n si
gnal
i
s
co
nve
r
ted
b
a
ck
to
t
he
time
domain by in
verse FF
T transfo
rm, and
thus the displacement si
gnal y(t) of time domain i
s
obtaine
d.
4. Acquisiti
on of Rela
tiv
e
Motion be
tw
e
e
n
T
w
o Si
de-B
y
-
Side
Ships
After shi
p
m
odel te
st, m
easure
d
d
a
ta
we
re
proce
s
sed to
obtai
n the
rel
a
tive motio
n
betwe
en two
side
-by-side
ship
s in wa
ves. This
se
ction de
scri
b
e
s the met
h
o
d
to pro
c
e
s
s the
measured da
ta of model test for relative motion bet
wee
n
two
sid
e
-by-sid
e
shi
p
s, and th
e key to
the method is the pro
c
essi
ng of accele
ration sig
nal.
4.1. EMD adaptiv
e
filter of acc
elera
t
ion data
In orde
r to get accu
rate di
splacement, a
c
celeration
s sign
al sho
u
ld
be adaptivel
y filtered
at first. If there are hi
gh fre
quen
cy
rand
o
m
noi
se
s o
r
some b
and
no
ise
s
, the a
cceleratio
n
sign
al
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Measured Da
ta Proce
s
sing
Method Fo
r Relati
ve Moti
ons …
(Ping-an Shi)
79
sho
u
ld be de
comp
osed by
EMD to filter the co
rre
s
p
o
n
d
ing freq
uen
cy band of noi
se
s. Otherwi
se
,
only the unavoidabl
e low-freque
ncy oscil
l
ation noise
n
eed to be filtered, whi
c
h
ca
n be achi
eve
d
by the summ
ation of all the IMF comp
o
nents b
u
t resi
due.
Becau
s
e
the
cal
c
ulatio
n of
a rel
a
tive mo
tion nee
d to
use t
w
o
ship’
s
six
kind
s
of motion,
inclu
d
ing
ea
ch ship’
s
surg
e, sway, he
a
v
e,roll,
pitch
and ya
w m
o
tions,
all of
which
ne
ed to
be
EMD ad
aptively filtered, b
u
t only the p
r
oce
s
sing
of
h
eave motio
n
of ship
A is
descri
bed
he
re,
other motio
n
s can be p
r
o
c
e
s
sed simil
a
rly
.
The mea
s
u
r
e
d
heave a
c
ce
leration d
a
ta of ship A is shown as Fig.
3.
Figure 3. Orig
inal mea
s
u
r
e
d
heave a
c
ce
leration
sign
a
l
In orde
r to
study the internal freq
uen
cy
cha
r
a
c
teri
sti
cs, the
heave
accele
ration
sign
al is
decompos
e
d into 10 IMF(c1,c
2,...,c
10) by EMD, as
is
s
h
own in Figure 4.
Figure 4. De
compo
s
ition of
heav
e accel
e
ration
sign
al
by EMD
It can b
e
see
n
from Fi
gu
re
4 that, c1,
c2, c3, a
nd
c4
co
rre
sp
ond t
o
the hig
h
fre
quen
cy
part of
the
a
c
celeration
si
gnal,
c5,
c6
and
c7
corre
s
po
nd to
the
ope
rating
freque
ncy
of the
sign
al, c8, c9
and c10 corre
s
po
nd to the low freq
uen
cy
of the signal,
and c1
0 is th
e trend term.
Figure 5 give
s the re
co
nst
r
ucte
d heave
accelera
tion
sign
al after EMD ada
ptive filtering,
whi
c
h represents the
re
su
l
t
s of low p
a
ss filtering, ban
d pa
ss
filteri
n
g and hi
gh p
a
ss filteri
ng fro
m
up to d
o
wn resp
ectively. Low
pa
ss filtering
is
th
e summation of c3, c4,
c5,
c6,
c7, c8, c9 and
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 73 – 82
80
c10, b
and p
a
ss filteri
ng i
s
the sum
m
atio
n of c3,
c4
, c5, c6, c7 and
c8, high
pa
ss filtering i
s
t
h
e
summ
ation of
c2, c3, c4, c5 and c6.
Figure 5. Re
constructe
d he
ave accele
rat
i
on sig
nal after EMD ad
ap
tive filtering
It can be
see
n
from Fig
u
re
5 that: the reco
nstructe
d
sign
al after lo
w freq
uen
cy filtering
has re
moved
high frequ
en
cy interefe
ren
c
es, b
u
t the
si
gnal i
s
n
o
t to
o sta
b
le; the
sign
al after hi
gh
pass filterin
g
has
high f
r
eq
uen
cy in
terfe
r
ence, but it is stable; th
e
si
gnal afte
r ba
ndpa
ss filteri
ng
not only
rem
o
ved hi
gh freque
ncy i
n
te
rfere
n
ce, but
also i
s
stab
le. So, EMD ba
sed
ad
ap
tive
filtering tech
n
o
logy ha
s go
od effect on ship’s motio
n
data filtering.
4.2. Processi
ng of Ac
cele
r
ation
After the me
asu
r
ed
a
ccel
e
ration
sig
nal
has
bee
n fil
t
ered
by EMD ad
aptive filter, it is
pro
c
e
s
sed to be tran
sform
ed into displ
a
ceme
nt. The pro
c
ed
ure to pro
c
e
ss a
c
cel
e
ration
sign
al
is
as foll
ows: (1
) To
tran
sfo
r
m the time
do
main a
c
cele
ration
signal
in
to frequ
en
cy
domain
si
gna
l by
Fast Fou
r
ie
r Tran
sfo
r
mati
on(F
FT); (2
) Integrate t
he f
r
equ
en
cy do
main sig
nal twice to obtai
n the
freque
ncy d
o
m
ain di
spla
cement; (3
) Inverse FFT
transfo
rmatio
n of the fre
quen
cy dom
ain
displ
a
cement
to obtain th
e
time domai
n
displ
a
cem
ent
; (4) EM
D a
d
aptive filterin
g of the resulting
displ
a
cement
.
Figure 6. Integration of He
ave accele
ra
t
i
on with an
d without EMD
adaptive filter
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Measured Da
ta Proce
s
sing
Method Fo
r Relati
ve Moti
ons …
(Ping-an Shi)
81
In orde
r to validate the efficiency of the
met
hod, the i
n
tegratio
n re
sults from a
cceleratio
n
to displa
cem
ent a
r
e
cal
c
ul
ated
with a
n
d
witho
u
t EMD ada
ptive filter
sep
a
rately.
Figure 6
give
s
the re
sult of i
n
tegratio
n of
one
ship’
s
h
e
a
ve ac
cel
e
rat
i
on
with and
wit
hout EM
D
adaptive filter. It
can b
e
se
en
from the re
sults that the EMD ada
pt
ive filter can e
ffectively eliminate the n
o
ise
embed
ed in the mea
s
u
r
ed
data.
4.3. Acquisition of Rela
tiv
e
Motion
After the
su
rge, sway
an
d he
ave a
cceleratio
n
sig
nal of
two
ship m
odel
s
h
a
ve be
en
transfo
rme
d
i
n
to displa
ce
ment a
c
cordi
ng to the a
b
o
ve-me
n
tione
d, and the
ro
ll, pitch an
d
yaw
sign
al have
b
een filtere
d
b
y
EMD ada
ptive filter,
the surge,
sway, heave, roll, pi
tch an
d yaw
of
two shi
p
mod
e
ls can be
su
bstituted into
(1),
(2)
a
nd (3
), and the lon
g
itudinal, late
ral and ve
rtical
relative moti
on can
be a
c
qui
red. Fi
gu
re 7
gives th
e longitu
dinal
, lateral a
nd
vertical
relati
ve
motion with
E
M
D
a
daptive filter,
whi
c
h a
r
e con
s
i
s
t
ent
with the
theo
retical
cal
c
ul
ation results
an
d
pra
c
tical situa
t
ion.
Figure 7. Vertical rel
a
tive motion with EM
D ada
ptive filter
5. Conclusio
n
s
The p
ape
r i
s
dedi
cated
to
pre
s
e
n
t an
effective met
hod to
pro
c
e
s
s mea
s
u
r
ed
data of
model te
st
wi
th co
ntact m
easure
m
ent
s
for the
relativ
e
motion
s
be
tween
two
si
de-by
-si
de
sh
ips
in wave
s. Th
e key to the
method i
s
the tran
sfor
m
a
tion of accel
e
ration
s to displacement
s, for
whi
c
h the
E
M
D a
daptive
filter and
freq
uen
cy dom
ai
n integ
r
aio
n
t
r
an
sform
ba
sed o
n
FFT
were
applie
d. The
relative motion obtaine
d throug
h this
method ag
re
e with the practical situati
on,
whi
c
h confirm
ed the effecti
v
eness of the
propo
se
d me
thod.
Referen
ces
[1]
F
ang MC
a
nd
Che
n
GR.
T
h
e
Rel
a
tive M
o
tio
n
a
nd W
a
v
e
El
evatio
n b
e
tw
ee
n T
w
o F
l
oati
n
g
Structure
s
in Waves
. Proceedings of
the Eleventh Internati
onal Offshore
and Polar E
ngineering C
onference.
Stavang
er, Nor
w
a
y
.
200
1: 361
-368.
[2]
Kim MS, Jeon
g HS, K
w
ak
H
W
, Kim BW
and Eom JK.
I
m
provem
ent Method on Offloading Oper
abilit
y
of Side-by-s
ide
Moored F
L
N
G
. Proceedin
g
s
of the
T
w
ent
y-Sec
o
n
d
Internatio
nal Sffsho
re and Po
l
a
r
Engi
neer
in
g C
onfere
n
ce. Rh
odes, Greece.
201
2: 921-
926.
[3]
Koo BJ, Kim MH.
Hydrodyn
a
mic Interactio
ns and R
e
lativ
e
Mo
tions of T
w
o F
l
oating Pl
atforms w
i
t
h
Mooring Lines in Side-
by
-side Offloading Pperation
. App
l
i
ed
Ocean Res
ear
ch. 2005; 2
7
: 292-3
10.
[4]
Che
n
GR, F
ang MC.
Hydrod
yna
m
ic Interac
t
ions betw
e
e
n
T
w
o Ships Advanci
ng in W
a
ves
. Ocean
Engi
neer
in
g. 2001; 28: 1
053-
107
8.
[5]
Godhav
en JM. Adaptiv
e T
unin
g
of Heave F
ilt
er
in Motio
n
Se
nsor. Proc. OCEANS; 1998: 1
74-1
78.
[6]
Liu J.C,
Xu Q
H
, Cha JX
.
M
e
thod
of Vi
bra
t
ion D
i
spl
a
ce
ment Me
asure
d
w
i
th Acceler
a
ti
on S
ensor
.
Modem ra
dar. 200
7; 29(5): 69
-71(In Ch
ines
e
)
.
[7]
F
l
andri
n
P, Ril
ling G, and G
onça
l
vés P. Empi
rica
l Mod
e
Decomp
ositio
n as a F
ilter Bank.
IEEE
SIGNAL PROCESSING LETTERS
. 2004; 1
1
(2):11
2-11
4.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 73 – 82
82
[8]
Huang
NE, Long SR, Wu MC, et al.
T
h
e E
m
pirica
l M
ode
D
e
co
mp
ositio
n
a
nd th
e H
i
l
bert
Spectru
m
fo
r
Nonl
in
ear a
nd
Nonstati
on
ary T
i
me Ser
i
es A
nalysis
. Proc.
R. Soc. Lond
o
n
. 1998, Ser
i
es
A (454): 903-
995.
[9]
Wu Z, Huang NE.
A
Study
of th
e C
har
a
c
teristics of
W
h
ite N
o
ise
Using
the
E
m
pirica
l M
o
d
e
Deco
mpositi
o
n
Method.
Proc. R. Soc. Lond. 200
4; Series A
(460): 159
7-1
611.
[10]
W
ang G, Ch
en
XY, Qi
ao F
L
.
On Intrinsic Mode Function
. A
d
vanc
es i
n
Ad
aptive
Data A
n
al
ysis. 2
0
1
0
;
2(3): 277-
29
3.
[11]
Cai F
., W
ang DY, Z
hang H,
Miao QM. Stud
y o
n
Sig
nal
Integratio
n Method B
a
sed
o
n
EMD Self-
ada
ptive F
ilter.
Journa
l of ship
mech
an
ics
. 20
07; 11(4): 5
28-
532 (In Ch
in
es
e)
Evaluation Warning : The document was created with Spire.PDF for Python.