TELKOM
NIKA
, Vol.12, No
.4, Dece
mbe
r
2014, pp. 98
5~9
9
6
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v12i4.367
985
Re
cei
v
ed Se
ptem
ber 3, 2014; Re
vi
sed
Octob
e
r 29, 2
014; Accepte
d
No
vem
ber
13, 2014
Information Support Technology of Ship Survey Based
on Case-based
Reasoning
Cao Jiy
i
n*
1
, Fan
Shidong
1
, Lu Wen
1
, Liu Haiy
un
2
1
Energ
y
an
d Po
w
e
r En
gin
eer
i
ng Co
lle
ge, W
uha
n Univ
ersit
y
of T
e
chnolo
g
y
, W
uha
n Chi
n
a
2
Metallurg
y a
n
d
Po
w
e
r En
gi
n
eeri
ng Co
lle
ge,
Hebe
i Unite
d
Univers
i
t
y
, T
angsha
n Chi
n
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: caoji
y
i
n
_
w
h
u
t
@
16
3.com
A
b
st
r
a
ct
Rece
ntly, the signific
anc
e of
ship i
n
specti
o
n
s has b
e
e
n
i
n
creas
ing
l
y re
cogn
i
z
e
d
bec
a
u
se se
a
poll
u
tio
n
a
nd s
a
fety pro
b
le
ms
are
occurri
ng
mor
e
a
nd mor
e
freq
uently. C
u
rrently, most ship ins
pectio
n
s
rely o
n
th
e ex
p
e
rie
n
ce
and
pr
ofessio
nal
kn
o
w
ledge
of th
e
w
o
rkers. Henc
e, it is
difficult f
o
r new
w
o
rker
s to
assess th
e sh
i
p
state i
n
th
e
ship
ins
pectio
n
s
.T
he pres
ent
prob
le
ms
are t
hat the
shi
p
i
n
spectio
n
tech
ni
cal
supp
ort leve
l
in Ch
in
a is n
o
t bal
ance
d
,
espec
ial
l
y as
to the curre
nt situatio
n w
i
th low
level,
po
or
techno
lo
gyin
u
nder-
deve
l
o
p
e
d
ar
eas. In
this
pa
per, th
e ca
se tech
nol
ogy
abo
ut the
case
-base
d
re
aso
n
i
n
g
to the ship
in
spectio
n
is pr
opos
ed. T
he
meth
ods
of normative i
n
sp
ection cas
e
r
epres
entati
on
are
discuss
ed.T
h
is
is consi
der
ed t
o
be th
e bas
is
of case-b
as
ed
reaso
n
in
g. T
h
e
n
the tertiary c
a
se structure w
i
th
the in
dex is
de
fined, i
n
w
h
ich
the
K-ne
arest
nei
ghb
or
meth
od to ca
lcul
ate
the si
mil
a
rity b
e
tw
een cases
w
as
used s
o
th
at the sh
ip
’
s
i
n
sp
ection
infor
m
ation c
an
be s
e
arche
d
effectiv
ely. In a
d
d
i
tio
n
, an
i
m
prov
e
d
retrievalstr
ateg
y of the ne
ar
est nei
ghb
or
meth
od
is
intr
oduc
ed. T
her
e
f
ore, in the
in
spectio
n
site,
the
insp
ectors ca
n
acq
u
ire
sup
p
o
rt infor
m
ation
by th
e cas
e
bases,
an
d th
en th
e n
e
w
ca
ses ar
e ca
lcul
ate
d
auto
m
atic
ally.
Further, a ship
inspecti
on c
a
s
e
mana
ge
ment
w
a
s introduce
d
to i
m
prov
e the stab
ility of th
e
system
. By carrying the case-bas
ed
reas
oning into an inspection, an
inspector can generat
e fault
infor
m
ati
on
an
d ins
pecti
on i
n
formatio
n
si
mp
l
y
and
eas
ily
. S
o
me ex
a
m
pl
es
of the or
gan
i
z
ation
an
d retrie
val
are show
n at the
en
d of the p
aper.
Ke
y
w
ords
:
shi
p
insp
ectio
n
, case-b
ased r
e
a
s
oni
ng, case re
trieval, K-ne
igh
bor metho
d
, ca
se ma
na
ge
me
nt
1. Introduc
tion
With the
rapi
d develo
p
me
nt of our cou
n
try's
ship
buil
d
ing a
nd
ship
tran
spo
r
t, the volum
e
of shi
p
inspection
business
will
al
so sy
nchronous grow [1],[2]. In
order to ensure
the safety of
navigation, jo
b security a
nd p
r
event t
he poll
u
tion
of the ma
rin
e
environme
n
t, it need
s
shi
p
insp
ectio
n
to
en
su
re th
at the
shi
p
h
a
ve a
goo
d
tech
nical
condition
[3]. Ship i
n
spe
c
tion
institution
s
can inspe
c
t the marin
e
ma
teri
als, me
ch
anical equip
m
ent and m
a
rine e
ngin
e
e
ring
facilities
according to the
provisi
o
ns of
state a
nd the rel
e
vant procedure,
so t
hat the
ship
will
meet the req
u
irem
ents of
the relevant i
n
ter
natio
nal
conve
n
tion
s and follo
w the national la
ws,
regul
ation
s
a
nd the tech
n
i
cal ind
e
xes
of inspe
c
tion
agen
cy rule
s [4]. Theref
ore, it is very
importa
nt to improve the q
uality of the ship inspe
c
tion
[5].
Curre
n
tly, shi
p
in
spe
c
tion
s are
po
orly
contro
lle
d, an
d their detail
s
rely o
n
the
e
m
piri
cal
knowledge of the pers
ons in charge.
Therefore,
the
business capability
and knowledge
cla
ssifi
cation
survey data
of
a Surveyor dire
ctly affects t
he q
uality and le
vel of the shi
p
insp
ectio
n
[6]. In this situat
ion, shi
p
in
sp
ec
tion
ability of the shi
p
m
anag
ement d
epartm
ent ve
ry
difficult to improve, resulting in low ship i
n
sp
ectio
n
technical su
ppo
rt level in China.
By the sup
p
o
rt of the
mode
rn
com
puter
te
ch
no
logy [7], network te
ch
nol
ogy [8],
informatio
n tech
nolo
g
y [9], ship inspe
c
tion te
chnol
ogy is al
so i
n
the ne
w chang
es. Sm
art,
efficient, an
d
reliabl
e i
s
the
goal
of mo
de
rn
ship
in
spe
c
tion te
ch
nol
ogy re
se
arch.
The te
ch
nolo
g
y
of case-b
ase
d
rea
s
oni
ng i
s
one effe
ct
ive method of solving this problem.
Re
cently, the case of surv
eyor usi
ng kn
owle
dge to solve the probl
em of ship in
spe
c
tion
is g
r
ad
ually i
n
crea
sing. T
h
e inspe
c
tion i
n
stitutio
n
s
oft
en
con
s
um
e
a large a
m
ou
nt of man
p
o
w
er,
material
and
pre
c
iou
s
ti
me re
so
urce
s when fa
ci
n
g
the ship i
n
sp
ectio
n
. T
herefo
r
e,
we
can
stand
ardi
ze
d
manag
e the
ship in
spe
c
tio
n
knowl
edg
e
and exp
e
ri
en
ce
of the exi
s
ting re
gulatio
ns,
and use
th
e case
-ba
s
e
d
re
aso
n
ing
me
chani
sm
to
gui
de solving
th
e sub
s
eq
uent
problem. It is of
great hel
p for inspe
c
tors to
redu
ce the rese
ar
ch time,
ensu
r
e the q
uality of inspe
c
tion.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 4, Dece
mb
er 201
4: 985
– 996
986
Curre
n
tly, th
e ca
se-ba
s
e
d
rea
s
oni
ng
has b
een u
s
ed in many other field
s
. A solution
retrieval
syst
em for expert
finding and
probl
em diag
nosi
s
by usin
g ca
se-ba
s
ed
reasonin
g
was
prop
osed [10
]. And a hybrid model was
develop
ed
by integrat-i
ng a ca
se-ba
s
ed
data clu
s
teri
ng
method
and
a fuzzy d
e
ci
sion
tree
for medi
cal
dat
a cl
assification [11]. In t
he
ship
field, an
automated
case
lea
r
nin
g
method fo
r
CBR-b
a
s
ed co
llision avoida
nce
sy
stems wa
s
introdu
ced
[12]. Howeve
r, the techn
o
lo-gy of ca
se
-based re
ason
ing ha
s not b
een u
s
ed for
ship in
sp
ectio
n
.
In this
pap
er,
we m
a
ke a
prelimina
r
y stu
d
y of
the
ca
se re
presentati
on of
shi
p
in
spectio
n
and th
e b
u
ild
issue
s
abo
ut the
ca
se
ba
se. In
additio
n
, the
ship
in
spe
c
tion
case fram
e mo
d
e
l is
con
s
tru
c
ted a
nd
the
ca
se retrieval strat
egy
by
the purpo
se of researchin
g the
ship in
sp
ecti
on
techni
cal
re
source
s i
s
di
scusse
d. Th
e last,
a
n
i
n
formatio
n tech
nolo
g
y to su
ppo
rt ship
insp
ectio
n
s,
whi
c
h u
s
e
s
the ca
se-ba
s
ed
reasoni
n
g
for ship
s, is co
m
p
reh
e
n
s
ively examined.
2. Case
Repr
esen
tatio
n
of Ship Inspe
c
tion
2.1 Cas
e
-b
as
ed Rea
s
onin
g
Technolog
y
O
v
er
v
i
e
w
CBR (Ca
s
e
-
based Rea
s
oning
) tech
n
o
logy origi
n
ates in the
United Sta
t
es Yale
university Ro
ger Sch
a
n
k
i
n
19
82 i
n
th
e
Dyna
mic de
scriptio
n
of the
Memo
ry, is a
ne
w
rise in
t
he
area of artifi
cial intellige
n
c
e, is an im
portant kno
w
ledge
-ba
s
e
d
probl
em solvi
ng and lea
r
n
i
ng
method. It is
to solve th
e
probl
em by
reusi
ng o
r
m
o
dify the sol
u
tion to the
pro
b
lem of
simil
a
r
before [13]. As we all kn
ow, huma
n
being
s always se
arch fro
m
their mem
o
rie
s
for sim
ila
r
probl
em
s tha
t
they have solved
s
u
c
c
es
s-fully in the pas
t to
fi
n
d
a solutio
n
, when they
are
confronted
wi
th a ne
w
prob
lem. Th
e
so
-
c
alled ex
p
e
ri
e
n
ce
is
st
o
r
ed
in CB
R
sy
st
e
m
in t
h
e
f
o
rm
of
ca
se
s.
Wh
e
n
en
cou
n
t
e
ri
ng ne
w p
r
o
b
lems,
CB
R
sy
st
em
s
s
ear
ch
ca
se
bas
e f
o
r
si
milar
experie
nces t
o
cou
n
t on.
Usi
ng
ca
se-b
ase
d
re
asoni
ng me
cha
n
isms to g
u
ide
the re
solutio
n
of shi
p
in
spectio
n
probl
em
s, ca
n re
du
ce te
sting time, imp
r
ove test e
ffici
ency a
nd a
ssure i
n
spectio
n
quality, so
that
insp
ect
i
o
n
t
a
sk
s ca
n be s
u
c
c
e
ssf
ully
c
o
mplet
e
d
by l
o
w te
ch
nolo
g
i
cal l
e
vel
su
rveyor. The
r
e
are
comm
only four cy
clical proce
s
se
s in CBR [14],
besi
des the
attempt in trying to build-up a
hig
h
quality ca
se
base [15] an
d rep
r
e
s
e
n
t ca
se
s: Re
tri
e
ving simila
r
ca
se
s; Re
usi
ng solution
s
of
simila
r ca
se
s;
Revisin
g
the prop
osed sol
u
tions; an
d Retaining the n
e
w case.
The ship inspectio
n
tech
nology ba
se
d on ca
se
-based re
aso
n
ing is
simil
a
r to other
method
s to solve the probl
em with the CBR, is sh
own
in figure 1.
A case is u
s
e
d
to rep
r
e
s
en
t the state de
scri
ption
of a probl
em an
d
its sol
u
tion st
rategy.
The pri
m
ary
missi
on of d
e
sig
n
ing a
Case
-ba
s
e
d
re
aso
n
ing te
ch
nology ship i
n
sp
ectio
n
ca
se
libra
ry is to
repre
s
e
n
t ship
insp
ectio
n
te
chni
cal i
n
formation a
s
th
e form
of case re
asona
bly and
effec
t
ively.
2.2 Cas
e
Tep
r
esen
ta
tion Metho
d
Ca
se
rep
r
e
s
e
n
tation mu
st
be b
a
se
d o
n
the vario
u
s e
x
isting
kno
w
l
edge
re
pre
s
e
n
tation,
almost all
exi
s
ting kn
owle
d
ge
rep
r
e
s
ent
ation can
be use
d
a
s
ca
se
rep
r
e
s
e
n
tation
meth
od
[1
6].
For the sam
e
ca
se, Ca
ses can be e
x
presse
d usi
ng different
method
s. Ho
wever, to sol
v
e a
probl
em, Different
rep
r
e
s
entation
s
ma
y produ
ce
co
mpletely dif
f
erent results. Therefo
r
e,
for
solving the
probl
em
s in different are
a
s, sel
e
ct
th
e appropri
a
te ca
se re
pre
s
entatio
n is very
importa
nt. A tech
nical ca
se
of ship in
spe
c
tion cont
ain
s
test items, in
spe
c
tion lo
cat
i
on, insp
ectio
n
stand
ard
s
a
n
d
so
on. The
r
efore, a
singl
e ca
se
of re
p
r
esentation i
s
difficult to m
eet the ne
ed
s. In
this p
ape
r,
obje
c
t-o
r
iente
d
technol
ogy
and
d
a
taba
se te
ch
nolo
g
y
com
b
ine
d
into th
e
ca
se
rep
r
e
s
entatio
n.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Inform
ation Suppo
rt Techn
o
log
y
of Ship Survey Ba
se
d on Ca
se
-ba
s
ed
Rea
s
o
n
in
g (Ca
o
Ji
yin
)
987
Figure 1. Basic step
s to sol
v
e the ship in
spe
c
tion p
r
ob
lem based on
CBR
So the adva
n
tage
s of re
pre
s
entatio
n
can b
e
u
s
ed.
And usin
g this meth
od can also
make goo
d use of
inhe
ritance
rel
a
tion
ship of
cla
sses, e
s
tabli
s
h
a hierarchy
betwe
en
ca
ses,
facilitate the
organi
zatio
n
and retri
e
val of case library, p
a
ckag
e attribute data
、
cas
e
retriev
a
l
、、
m
o
dification
met
hod
s
p
reservation metho
d
s a
nd
so
on
. Dynami
c
ally con
s
truct
ca
se-
objects when
the system i
s
runni
ng[17],[18].
2.3
Cas
e
Re
presen
ta
tion
of Ship Inspection
Ca
se
conte
n
t usu
a
lly con
s
ists
of thre
e p
a
rt
s
<De
s
crip
tion of the p
r
oblem,
De
scription of
the solutio
n
, De
scription of
the effect>
(1)
De
scriptio
n of the probl
em:
De
scribe
d the state of environ
m
ent
whe
n
the pro
b
lem occu
rs.
(2)
De
scriptio
n of the soluti
on: Illu
strate
s the solution f
o
r the proble
m
s.
(3)
De
scriptio
n of the effect: asse
ss
ment
and sum
m
ary of treatment plan.
So, case of ship inspe
c
tio
n
that based
on ca
se
-b
ase
d
rea
s
o
n
ing
can b
e
defin
ed as
a
quad
rupl
e:
,,
,
CD
S
R
T
(1)
In whic
h,
12
3
4
,,,
.
.
.
Dd
d
d
d
is a
non-empty fin
i
te set, it re
prese
n
t de
scrip
t
ion
of shi
p
in
spe
c
tion, in
cludi
n
g
case nu
mb
er, ship
para
m
eters, ship i
n
sp
ectio
n
p
r
o
c
e
s
ses, type
of
ship's inspect
i
on
;
12
3
4
,
,
,
...
Ss
s
s
s
is a non-empty finite set, it represe
n
t feature set
of
ship
in
spe
c
ti
on
ca
se;
R i
s
con
c
lu
sio
n
informatio
n o
f
ca
se; T
is feedb
ack
of case. S
o
, a
ship
insp
ectio
n
ca
se can defin
e
object
s
of structure sh
own
in Table 1.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 4, Dece
mb
er 201
4: 985
– 996
988
Table 1. Stru
cture
s
of Ship
Inspe
c
tion
Ca
se
Structure
Content
Description of ship
inspection case
case number
ship parameters
ship inspection p
r
ocesses
…
t
y
pe of ship insp
ection
feature set of shi
p
inspection case
eigenvector
w
e
ight vector
conclusion
information of ca
se
conclusions
IETM
cer
t
ificates
…
feedback of case
evaluation of results and
summary
case study
quality
the number
of su
ccessful
match
r
eusability
Ca
se for
shi
p
insp
ectio
n
fe
ature
set is
d
e
scri
b
ed in
d
e
tail, ship in
spectio
n
Case
feature
set is a no
n-e
m
pty finite set, repre
s
entin
g the
value, prope
rty and d
a
ta type of each featu
r
e.
(1) Ship in
sp
ection
Ca
se
Eigenvecto
rs
Ship inspe
c
tion ca
se
eige
nvectors me
a
n
s all ei
genv
alue
s co
mpo
s
ed in
a ce
rt
ain order
vector, after
para
m
eteri
z
e
d
treatme
nt. The mai
n
fe
ature
s
in
clud
e: ship i
n
spe
c
tion
reg
u
latory
requi
rem
ents,
previo
us
su
rveyor in
spe
c
ti
on data,
environmental
pa
rameters
et
c. Before ca
rryi
n
g
out ca
se
re
asoning
,
ch
ara
c
teri
stic val
u
e
s
mu
st be
expre
s
sed a
s
a
rang
e o
r
limit
ed set of valu
es.
Ship insp
ecti
on feature ve
ctor i
s
expre
s
sed a
s
:
12
1
2
...,
,...,
,
1
,
,
1
,
ii
A
A
A
Ak
An
i
m
k
n
,,
,
i
A
mean
s shi
p
insp
ectio
n
eig
envecto
rs in
the i-th ca
se,
i
Ak
mean
s the ei
genvalu
e
s th
at come
from the k-th indic
a
tors
after
the
pa
ra
meteri
zed
tre
a
tment in th
e i-th
ca
se.
As the
statut
ory
insp
ectio
n
ch
ara
c
teri
stics t
o
die
s
el,
of which
sh
ip i
n
sp
ection
characteristics gen
e
r
al p
e
rfo
r
ma
n
c
e
as thre
e types:
a.
Logi
c type. State exists o
n
l
y True o
r
Fa
ls
e.0 me
an
s the si
gn do
es not app
ear,1
mean
s th
e
sign
ap
pea
r.
Like
the
safe
ty interlo
c
k b
e
twee
n p
o
we
r tra
n
sfe
r
dev
ice
and
sta
r
ti
ng d
e
vice
in
diesel engin
e
:
0
()
1
i
i
i
a
T
RUE
Va
a
F
ALSE
(2)
In which,
a
i
is
the i-th attribute,
i
Va
is the value of it.
b. Nume
ric.
Quantitativ
e data, Such a
s
po
wer
、、
sp
e
e
d
c
ommuta
tion time of a diesel engi
ne,
ship in
sp
ecti
on pro
c
e
s
s is always con
s
ide
r
ed that
the value is i
n
a rang
e of
rea
s
ona
ble,
regul
ation
s
al
so p
r
ovide
s
i
n
a ra
nge. F
o
r this fe
ature, Ca
se cha
r
acteri
stic val
ue is e
qual t
o
the actual m
easure
m
ent value of shi
p
insp
ecti
o
n
dat
a. Ship inspe
c
tion case attribute
s
hav
e
different dim
e
nsio
ns
and o
r
de
rs
of mag
n
itude in p
r
a
c
tical
appli
c
at
ions,
so it mu
st be u
n
ified
before
we ne
ed apply the
s
e prop
ertie
s
.
1
,1
-
(
,
)
1
-
(
,
)
n
ii
i
i
SI
M
X
Y
D
I
S
T
X
Y
W
D
X
Y
(3)
In which,
i
V
is prope
rty value
s
,
12
si
,
.
.
.
i
mVV
V
is attribute
values after
dimen
s
ionl
ess
transfo
rme
d
, and
12
si
,
.
.
.
1
,
1
i
mV
V
V
.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Inform
ation Suppo
rt Techn
o
log
y
of Ship Survey Ba
se
d on Ca
se
-ba
s
ed
Rea
s
o
n
in
g (Ca
o
Ji
yin
)
989
c. With a text description
of the disord
er
enu
meration data. For example, bo
iler feed water
pressure: “too large, norm
al, too small”
etc. In
order to facilitate data
processi
ng, we required
to digitize them. We
can establish the corr
esponding index table, scilicet, create a index
table of which pro
p
e
r
ty an
d Its index va
lue are corre
s
po
ndin
g
. Ind
e
x values
are
rep
r
e
s
ente
d
with intege
rs 0, 1, 2 etc. Particula
r
ly, prov
ide
s
th
at all normal
attribute values
can
be
rep
r
e
s
ente
d
with 0.Differe
nt situation
s
have di
ffere
n
t
attribute ind
i
cate
s, you can ma
ke the
approp
riate chang
es, table
2
sho
w
s the i
ndex table.
Table 2. Di
so
rde
r
ed
Data Indexing Ta
bl
e
Index Propert
y
0 Normal
1
Too high or
too l
a
rge etc.
2
Too lo
w
or
too s
m
all etc.
(2) Ca
se
c
har
acteri
st
ics we
ight vector
Different
ca
se ch
ara
c
te
ristic attribute
s
in ship i
n
spe
c
tion fun
c
tion
is differe
nt in shi
p
insp
ectio
n
proce
s
s. By introdu
cing
t
he
weig
ht
v
e
ct
or
,
it
makes t
h
e
result
s of
si
milarit
y
cal
c
ul
at
ion
more rea
s
on
able [19].Th
e
traditional
methods
fo
r determi
nin
g
a feature
weight
s co
ntain
con
s
ultin
g
experts, d
o
mai
n
kno
w
le
dge,
survey st
ati
s
tics, etc. The
s
e meth
ods
are
simple, fast
but due to
over-relia
nce on
subj
ective ju
dgment a
nd
experie
nce, it is sometime
s difficult to g
e
t a
rea
s
on
able
solution
ca
se.
No
n-traditio
nal meth
od
s co
ntain g
e
n
e
tic alg
o
rith
ms
(GA), An
alytic
Hierarchy Proce
s
s (AHP
),
which
overcomes th
e sh
ortco
m
ing
s
o
f
traditional
method
s. But the
algorith
m
is
compl
e
x and
it is difficult
to ac
hieve. K
nowl
edge in
this field will
adopt la
ws a
nd
roug
h
set the
o
ry combin
ed
method
to e
x
tract t
he wei
ghts. Ca
se
fe
ature wei
ght vector ca
n
b
e
expre
s
sed
as: {Ca
s
e
featu
r
e
weig
ht, nu
mber weight
value}
t
w
o pa
rts. Due
to space
limitatio
n
s
not descri
bed
in detail here
.
3. Organizati
on and Re
tri
e
v
a
l of Ship
Inspec
tion Case Base
3.1. Ship Inspection
Cas
e
Bas
e
Orga
nization
Orga
nize the index ca
se i
s
conveni
ent for the
ret
r
ieval. Acco
rdin
g
to the chara
c
teri
stics
of the ship in
spe
c
tion, the
retrieval
strat
egy co
m
b
inin
g ship i
n
spe
c
tion pro
c
e
s
s with the ne
arest
neigh
bor
alg
o
rithm i
s
taken in this
pa
per. When t
he case ba
se is
small, a
nd the any t
w
o
prop
ertie
s
are mutually i
n
depe
ndent, t
he
simple
st
a
nd n
earest
n
e
ighb
or
algo
rithm is th
e m
o
st
effective retri
e
val techni
q
ues. But wh
en nea
re
st n
e
ighb
or meth
od is u
s
e
d
to cal
c
ulate t
he
simila
rity, ea
ch
ca
se
ha
s
to be
cal
c
ul
a
t
ed ag
ain.
A
s
t
he
ca
se
ba
se i
n
c
r
e
a
se
s,
it
will
def
init
ely
lowe
r
ca
se
retrieval effici
ency, it is ne
ce
ssary
to
organi
ze the
case
ba
se.
Now,
shippi
ng
test
case library
will be divi
ded into three
classes,
the inspection
pr
ocesses
of whi
c
h are
different.
The con
s
tru
c
tion insp
ectio
n
pro
c
e
s
s of
a 2700
0DWT multi-pu
rpo
s
e
s
vessel
was ta
ken a
s
an
example, the ca
se ba
se
st
ructure sh
own
in Figure 2.
E
a
ch t
y
pe
of
spe
c
if
ic
c
a
s
e
s
con
s
t
i
t
u
t
e
s
a sm
alle
r
spe
c
if
ic
ca
se
libr
a
ry
,
nam
ely
a
sp
ecif
i
c
layer. Sele
cting a
rep
r
e
s
e
n
tative of ca
se in
the
spe
c
ific
ca
se
as the ind
e
x of the catego
ry o
f
ca
se
s, all
cla
s
s represent
atives
con
s
titute a
re
pre
s
entative case
ba
se,
whi
c
h
belo
n
g
s
to
an
abstractio
n
la
yer an
d is
a
typical, re
pre
s
entativ
e
ca
se, on be
half
of a cl
ass
ca
se in
ret
r
ieva
l.
Process-l
e
vel
Ca
se
Lib
r
ary
refe
rs to a
case
ba
se
un
der
a in
sp
ecti
on p
r
o
c
e
ss.
Whe
n
sea
r
ch
ing
it, the first st
ep is to
dete
r
mine which o
ne p
r
oces
s i
n
a inspe
c
tion
mode, the
n
to find the m
o
st
simila
r
rep
r
e
s
entative case
on
re
pre
s
e
n
tative ca
se
ba
se
(thi
s
step i
s
e
quivale
nt to ne
w i
s
sue
s
of
cla
ssif
i
cat
i
on,
t
o
jud
g
ing
w
h
ich
c
a
t
ego
ry
of
cas
e
s
it
belongs
to), finally to
cond
uct further retrie
val
in the most si
milar case co
rre
sp
ondi
ng to the ca
se of that kind.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 4, Dece
mb
er 201
4: 985
– 996
990
Figure 2. Ship insp
ectio
n
case b
a
se stru
cture
3.2.Similar Design of Cas
e
Bas
e
The fun
dam
e
n
tal pu
rpo
s
e
of invoki
ng
case
s i
s
to
co
mpare the
si
milarity bet
ween th
e
ca
se in ba
se
case and th
e pro
b
lem
s
to be solve
d
and find the
most simil
a
r
ca
se
s. Thu
s
, the
simila
rity is the basi
s
for
ca
lling ca
se.
This pa
per t
a
ke
s K-ne
arest neigh
bou
r method se
arching meth
od to calcul
ate the
simila
rity betwee
n
ca
se
s [20]. Whe
r
e the nea
re
st neighb
or meth
od (Nea
re
st Neig
hbo
r Met
hod,
NNM
) is a
sp
ecial
ca
se of K-nea
re
st nei
ghbo
r metho
d
(K=1
) [18].
Suppo
sing
12
X{
}
n
XX
X
,,
,
,
1
i
X
in
is its i-th eigen
value.
i
W
is its
weight. X
is a
poi
nt of
n-dim
e
n
s
iona
l feature
spa
c
e
12
D
n
DD
D
.
ii
X
D
, As
the X,
Y in
s
p
ac
e D,
X, Y distance
on D is follo
wed:
1
(,
)
(
,
)
n
r
r
ii
i
i
DIS
T
X
Y
W
D
X
Y
(4)
i i
i
i i
i
i
0
W
hen
D
i
s d
i
sc
r
e
te
and X
Y
(
,
)
1
W
h
e
n
D
i
s
di
s
c
r
e
t
e
and X
Y
W
hen D
i
s
c
onti
nued
ma
x
m
i
n
ii
ii
ii
DX
Y
XY
,
,
(5)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Inform
ation Suppo
rt Techn
o
log
y
of Ship Survey Ba
se
d on Ca
se
-ba
s
ed
Rea
s
o
n
in
g (Ca
o
Ji
yin
)
991
Whe
r
e:
ma
x
i
and
mi
n
i
denote the maximum an
d minimum
values of th
e prop
erty
respec
tively,
when r is
2 in the formula (4),
(,
)
DIST
X
Y
is the Euclidean di
stan
ce.
Whe
n
the
distance
bet
wee
n
the
ca
se i
s
defin
e
d
, the
n
the
simila
rity betwe
en
ca
se
s
ca
n
be define
d
as:
1
,1
-
(
,
)
1
-
(
,
)
n
ii
i
i
SI
M
X
Y
D
I
S
T
X
Y
W
D
X
Y
(6)
3.3. Search
Strateg
y
of
Cas
e
Ca
se ret
r
ieva
l is not only a key step to
achi
eve
CBR, but also t
he co
re of CBR expert
system. T
he
main p
u
rp
ose is to
retri
e
ve a g
r
ou
p of
simila
r cases as little a
s
p
o
ssible f
r
om t
h
e
ca
se li
brary b
a
se
d o
n
the
n
e
w
problem
d
e
finition a
nd
descri
p
tion,
which
have
referen
c
e
value
to
the probl
em
as the ba
si
s for solving
new p
r
oble
m
s. Typically
, case
kno
w
l
edge
sea
r
chi
ng
strategy h
a
s nearest nei
ghbo
r strate
gy, inducti
ve
rea
s
oni
ng strategie
s
, kn
owle
dge g
u
i
d
ing
strategi
es, te
mplates retri
e
val strategi
es
and
so
o
n
. Wh
ether
sea
r
ch
strate
gy sel
e
ction
is
approp
riate, the high
-spee
d and effici
en
t completio
n
of case ret
r
ie
val performa
n
ce h
a
s
a direct
impact on problem solving.
The detaile
d step
s to retrie
ve in the case of a class li
bra
r
y are follo
wed:
(1)
Comp
ari
ng n
e
w i
s
sue
with
the m ca
se i
n
the
library (comp
a
ri
so
n o
f
n con
d
itiona
l attribute),
the comp
ared
result
s are saved as a ma
trix of the form:
11
12
1
12
n
ij
mm
m
n
(7)
Whe
r
e i
= 1
,
2 ……m, j = 1,2…
… n. i denote
s
the i-th
ca
se,
j represents the j-th
con
d
itional at
tributes,
ij
is th
e local
simila
rity compari
n
g
new p
r
obl
em
s with the i
-
th
ca
se of the
j-th con
d
ition
attributes
(2)
The wei
ght vector
T
12
n
ww
w
is multiplied by the matrix
(
j
w
is the j condition at
tribute
weig
hts), th
e
result of
which is
T
12
n
, namely
the overall si
milarity betwe
en the
ne
w
ca
se
s and e
a
c
h of the m case. Spe
c
ific
formula i
s
as
follows:
11
11
1
2
1
22
12
n
ij
mm
m
n
nn
w
w
w
(8)
(3)
Selecting
a o
r
k
ca
se
s
wit
h
the maxim
u
m val
ue
of a degree of
sim
ilarity
as a se
arch
result.
Retrieve
d ca
se may
b
e
positive ca
se
scen
ar
io
o
r
co
unte
r
exa
m
ple. Po
sitive exam
ples
provide
expe
rien
ce, while
cou
n
ter-exam
ples
provide
a le
sson, whi
c
h
ca
n avoid
rep
eating
past mi
stake
s
.
j
w
, as the val
u
e of th
e p
r
o
perty ri
ghts
of t
he ca
se, reflect
s
th
e
d
egre
e
of
infl
uen
ce
whi
c
h th
e p
r
operty
ha
s o
n
the
re
sult
s .the rea
s
on
able
wei
ghts ha
s g
r
e
a
t i
n
feren
c
e
on
the
accuracy
of t
he
re
sults of
the
ca
se. T
here
a
r
e
usu
a
lly two
setti
ng m
e
thod
s:
one
is for the
surveyo
r
to
set the level of
experie
nce, whi
c
h
h
a
s
a
high d
egree
of subj
ectivity; another way is
the analysi
s
a
nd cal
c
ul
ation
of existing histor
i
c
al data t
o
get derived
weig
hts obj
ectively.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 4, Dece
mb
er 201
4: 985
– 996
992
In orde
r to a
c
hieve the
o
b
jective
s
, na
mely
rapid
retrieval of si
milar cases
of current
ca
se
s and
o
f
the matchi
ng ca
se
as
little as po
ssible, this
p
aper
esta
blished a
ship
pi
ng
insp
ectio
n
ca
se ret
r
ieval m
odel, sh
own in Figure 3.
The sp
ecifi
c
impleme
n
tatio
n
pro
c
e
ss i
s
as follo
ws:
(1)
Jud
ge
ca
se
s in the p
r
o
c
e
ss
und
er
so
me ki
nd of in
spe
c
tion
s th
rough th
e hu
man-com
pute
r
intera
ction int
e
rface.
(2)
Find th
e m
o
st
similar
ca
se i
n
th
e
re
pre
s
entative
ca
se
ba
se
u
s
ing
the
nea
rest
nei
ghb
or
method (K = 1).
Th
e
repre
s
entative ca
se
is putt
ed
fo
rwa
r
d with ne
w ca
ses, and
the
rest
a
r
e
filtered out. T
h
is ste
p
is e
q
u
ivalent to cl
assi
fy new
ca
se
s whi
c
h
bel
ong to the re
pre
s
entative
ca
se.
(3)
Put forward K neighbo
r me
thod for furth
e
r retri
e
val in a related sp
ecific
case to find the most
simila
r ca
se.
In the la
st two
stage
s, set two
simila
rity thre
shol
d respectively,
11
01
)
(
,
22
(0
1
)
, only a ca
se simila
rity whi
c
h i
s
gre
a
ter than th
e
corre
s
p
ondi
ng thre
sh
old
s
is
c
h
os
en . The two thres
holds
c
a
n artific
i
ally s
e
t.
Figure 3. Retrieval model
4. Case Ma
n
a
gemen
t
of
Ship Inspection
CBR-ba
se
d ship in
spe
c
tio
n
dep
end
s o
n
not only th
e pe
rform
ance of ca
se
ba
se
ca
se
rep
r
e
s
entatio
n, re
asonin
g
and
retriev
a
l st
ra
tegi
es,
but
also t
he
ca
se
ma
nagem
ent
a
n
d
maintena
nce. So the CBR ca
se ba
se m
anag
ement system is an in
disp
en
sable
part.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Inform
ation Suppo
rt Techn
o
log
y
of Ship Survey Ba
se
d on Ca
se
-ba
s
ed
Rea
s
o
n
in
g (Ca
o
Ji
yin
)
993
Figure 4. Ship insp
ectio
n
case d
a
taba
se
manage
ment
4.1. Unit Ca
s
e
manageme
nt
(1)
Adding ne
w
ca
se is the case b
a
se passive lear
ning
pro
c
e
ss. In o
r
de
r to keep
rich lib
ra
ry of
the ca
se, in a
ddition to exp
e
rt or p
r
ofe
ssional
o
r
ga
nization su
rveyo
r
ca
n get a
c
tual ca
se
s o
r
data sto
r
ag
e
con
s
oli
dation,
local
ship in
spe
c
tion p
e
rsonnel
sho
u
ld
also e
s
tabli
s
h ince
ntive
mech
ani
sm
s
to encou
rage
their p
a
rtici
p
ation in
the
work, to m
obi
lize the
enthu
sia
s
m of the
entire ind
u
st
ry maximize the wealth of case b
a
se.
(2)
Ca
se
co
rre
ct
ed. Ca
se
am
endme
n
ts in
clude am
end
m
ents to th
e co
ntent of the
case
and
ca
se
st
ru
ct
ur
e
o
p
t
i
mizat
i
on.
Ca
se co
nt
ent
correctio
n
refers to
revise
ch
ara
c
te
risti
c
attrib
utes,
attribute wei
g
hts, and tre
a
tment option
s
of ship in
spe
c
tion ca
se.
(3)
Ca
se del
ete
d
. Case ba
se cap
a
city is not better
whe
n
it is bi
gger. In the
prog
re
ss that
appli
c
ation
b
a
se
d on
CB
R
ca
se b
a
se ship in
sp
e
c
tion, the
r
e
may be
so
me cases of
mismat
che
d
or sl
ave. The
r
ef
ore, the knowl
edge
ba
se shoul
d be
streamli
ned
on a re
gula
r
basi
s
, and
make the n
e
ce
ssary del
etion und
er
the guida
nce
and parti
ci
pation of the
profe
ssi
onal
s.
4.2. Mainten
a
nce o
f
Ship Inspection
Cas
e
Librar
y
In orde
r to m
a
intain the
efficien
cy of re
as
o
n
ing and inferen
c
e accura
cy
of
the results,
ca
se
b
a
se sh
ould
b
e
controlled
at a certain
scal
e,
an
d the p
o
tenti
a
lly misle
adin
g
case
s
(noi
se)
on
the ca
se b
a
se sh
ould b
e
delete
d
o
r
update
d
.
An
d
timely optimi
z
e the
organi
zation
al st
ru
cture
and sto
r
ag
e structu
r
e
s
of case lib
ra
ry.
5. Examplesand Re
sults
5.1. Example of Cas
e
Ba
s
e
Organi
zati
on
The fault in the mari
ne en
gine fuel sy
stem and
cooli
ng syste
m
occurs fre
que
ntly, so it
use
s
the two
parts of a fault to analyse. The
insp
ection fault case
s present
s as a form
o
f
deci
s
io
n tabl
e, then use
s
ro
ugh
set
theory to
calcul
ate. The
deci
s
ion ta
ble co
ntain
s
six
con
d
ition
s
p
r
opertie
s
: to
rq
ue, temp
erat
ure, th
e th
rot
t
le ope
ning,
air flo
w
rate,
rotational
spe
ed
and pul
se
wi
dth. Deci
sio
n
prop
ertie
s
a
r
e fault ty
pes, inclu
d
ing
six kind
s of
failure
s: normal,
cooli
ng wate
r temperatu
r
e
is too high, a cylinde
r
ha
s no inje
ct
ion
circ
uit
,
t
o
rqu
e
sen
s
o
r
cir
c
uit,
temperature
sen
s
o
r
ci
rcuit, air flow se
n
s
or i
s
dam
ag
ed, re
spe
c
tively corre
s
po
n
d
ing to 0,1,2,3,4,
and 5.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 12, No. 4, Dece
mb
er 201
4: 985
– 996
994
Table 3. Engi
ne Insp
ectio
n
Failure
Ca
se
Case
number
Torque
(
V
)
Temper
ature
(
V
)
Throttle
percentage
(
V
)
Air
flow
rate
(
V
)
Rotational
Speed
(
r/mi
n
)
Pulse
wi
d
t
h
(
ms
)
Fault
t
y
pes
1 0.373
0.282
0.943
1.406
1317.6
4.49
2
2 0.31
0.83
1.07
1.625
1797.4
4.136
0
3 0.952
0.231
1.641
2.403
1855.6
7.638
0
4 3.179
0.498
2.072
1.676
1255.0
5.857
3
5 0.419
0.572
0.727
3.747
687.4
4.5
5
6 0.483
4.957
0.942
1.396
1053.9
5.224
4
7 1.26
0.055
1.726
2.314
1411.4
9.376
1
8 0.84
0.154
1.626
2.285
1403.3
8.8
2
9 0.637
0.389
1.223
1.821
1350.986
6.864
0
10 0.415
0.943
0.857
1.26
1123.007
4.288
0
11 1.575
0.077
0.938
2.754
1760.563
11.068
1
12 3.194
0.97
1.38
2.009
1210.38
7.888
3
13 0.39
0.665
0.727
3.767
625.904
4.5
5
Table 4.Simil
a
rity Matrix of Case 4
{1}
{2}
{3}
{7}
{8}
{9}
{10}
{11}
{1}
1
0.8924
0.6449
0.5019
0.5675
0.8483
0.8595
0.3286
{2}
0.8924
1
0.6230
0.4687
0.5357
0.8212
0.9152
0.3024
{3}
0.6449
0.6230
1
0.8293
0.8882
0.7925
0.5851
0.6724
{7}
0.5019
0.4687
0.8293
1
0.9108
0.6440
0.4442
0.8247
{8}
0.5675
0.5357
0.8882
0.9108
1
0.7103
0.5089
0.7522
{9}
0.8483
0.8212
0.7925
0.6440
0.7103
1
0.7828
0.4743
{10}
0.8595
0.9152
0.5851
0.4442
0.5089
0.7828
1
0.2775
{11}
0.3286
0.3024
0.6724
0.8247
0.7522
0.4743
0.2775
1
Make
a cl
ust
e
ring
analy
s
i
s
of the
s
e case
s, u
s
e M
A
TLAB for computing, a
simila
rity
threshold val
u
e
=0
.
7
. The final
clu
s
terin
g
tre
e
is sh
own in Figure 4.
Figure 5. Clu
s
terin
g
Ttree
of Cast Ba
se
So, the 4 s
pec
ific
c
a
ses
are put forwar
d: CASE1={6}, CASE2=
{
5,13}, CASE3=
{
4,12},
CASE4= {1,
2
,3,7,8,9,10,11}. Tak
i
ng
an exam
ple of the
CA
SE4=
{1,2,3,7,8,9,10,11},
the
Similarity matrix of case is
sho
w
n in tabl
e 4.
The sum
of
t
he similarity ca
se 1
to with
othe
r ca
se
s:
L
=
0.8
934
+ 0.643
9 + 0.5018
+
0.5675
+
0.8
483
+ 0.8
595
+
0.3286
=4.
6430. In th
e
same
way, the
sum
of
ca
se
3, 7, 8, 9, 1
0
, 11
to other
ca
se
similarity a
r
e
:
4.5594, 4.6
232,
4.873
4, 5.0733, 4.3
7
32, 3.
6319. S
o
, ca
se 9 i
s
the
Evaluation Warning : The document was created with Spire.PDF for Python.