TELKOM
NIKA Indonesia
n
Journal of
Electrical En
gineering
Vol.12, No.7, July 201
4, pp
. 5575 ~ 55
8
4
DOI: 10.115
9
1
/telkomni
ka.
v
12i7.537
4
5575
Re
cei
v
ed
De
cem
ber 1
5
, 2013; Re
vi
sed
Jan
uar
y 11, 2
014; Accepte
d
February 8,
2014
Construction of Geological Knowledge-based
Systems
of Railway Route Select
ion
Lv Xikui*
1,
2
, ZhouXiaopin
g
1
, He Bin
1
1
School of T
r
affic and T
r
ansportation, Shi
jia
zhua
ng
T
i
eDao
Universit
y
, 05
004
3, Shiji
azh
uan
g Chi
n
a
2
T
r
affic Safeties and Co
ntrol L
ab of Heb
e
i Pr
ovinc
e
Chi
na, 050
04
3, Shiji
a
z
hua
ng, Ch
ina
*Co
rre
sp
ondi
ng autho
r, e-mail: lvxikui_tdu@163.
com
A
b
st
r
a
ct
Accordi
ng to
the
divers
ity of
ge
olo
g
ic
al c
o
nditi
ons
invo
lv
ed
in r
a
ilw
ay
l
o
catio
n
d
e
si
gn
, different
types of g
eol
o
g
ical
ob
jects
w
e
re classifi
ed
and
in
dexe
d
, and pro
pose
d
the mod
e
li
ng meth
od bas
ed
on
mu
ltipl
e
su
b-b
a
ses. It use
d
the co
mputer'
s
intern
al
and
e
x
ternal r
epr
ese
n
tation
in
ge
ol
ogic
a
l k
now
led
g
e
repres
entati
on.
In computer e
x
ternal
, it used
object-ori
ente
d
class-ru
l
e
s know
led
ge repr
esentati
on
mo
de
l
,
add
ed cre
d
i
b
il
ity to the ob
j
e
ct-orie
n
ted k
now
led
ge r
epr
esentati
on
mo
del, a
nd th
en
achi
eved f
u
zz
y
repres
entati
o
n
of know
le
dg
e. In
ord
e
r
to achi
eve v
i
sual r
epres
en
tation of
externa
l
know
le
d
g
e
repres
entati
on,
text, imag
es a
nd thre
e-di
men
s
ion
a
l virt
u
a
l r
eality tec
hno
lo
gy w
e
re use
d
. Accordi
ng to t
h
e
ambi
guity of geol
ogic
a
l know
led
ge,
it reali
z
ed unc
ertainty
reas
oni
ng
of
t
he part bas
ed on
r
u
le
co
nd
iti
o
n
and co
ncl
u
sio
n
s
.
Ke
y
w
ords
: ge
olo
g
ica
l
know
le
dge b
a
se, rai
l
w
a
y r
oute sel
e
cti
on, reaso
n
i
ng
construct
Copy
right
©
2014 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introducti
on
De
cad
e
s of e
ngine
erin
g practice proved
that
it is necessary for
co
nstru
c
ting
a mode
rn
railway with
high q
uality to select
rail
way lines rea
s
onably in
rail
way
con
s
tru
c
tion a
c
cordi
n
g to
geolo
g
ical
co
ndition
s [1].
In mountai
no
us
are
a
s an
d tho
s
e
are
a
s
with
compl
e
x engin
e
e
r
i
n
g
geolo
g
ical
co
ndition
s, engi
neeri
ng
geol
ogy line
sel
e
ct
ion plays a
n
impo
rtant role
in
th
e rail
way
line sel
e
ctio
n
.
How to
sel
e
ct a rea
s
on
able rout
e a
c
cordi
ng to t
he ge
ologi
cal
con
d
itions?
In
addition to d
epen
d on re
gular route d
e
sig
n
st
and
a
r
ds, it depen
d more on route desi
gne
rs’
experie
nce [2]. But the
experie
nce a
nd kn
owl
edg
e can
not be
got overnig
h
t. Throug
h
the
establi
s
hm
en
t of geological kno
w
led
g
e
base of ro
ut
e sele
ction, it offers geol
o
g
ical
kno
w
led
g
e
requi
re
d to
ro
ute de
sig
ners, automati
c
all
y
sea
r
ch
e
s
re
lated lin
e-sel
e
ction
ge
olog
ical
kn
owl
edg
e
in the line-sel
e
ction p
r
o
c
e
s
s, and the
n
p
r
ovide
s
en
gin
eers re
al-tim
e help a
nd re
feren
c
e. It also
provide
s
ge
o
l
ogical ba
sis for railway line sel
e
ctio
n
by reasonin
g
and ma
ke
s it possible
for
building a full
rang
e of intelligent route
se
lection g
eolo
g
ical e
n
viron
m
ent for rail
way line sele
cti
on
and a
c
hievin
g remote
sen
s
ing g
eologi
cal route sele
ction in three
-
d
i
mensi
onal e
n
vironm
ent.
2. Design of
Kno
w
l
e
dge
Bas
e
Structu
r
e of kno
w
le
dge
b
a
se
play
s a
n
important
role in kn
owl
edge ba
se it
self. An
inco
nsi
s
tent
or in
co
mplet
e
kno
w
led
g
e
ba
se
co
ul
d greatly red
u
ce
the efficien
cy
of rea
s
o
n
ing.
Expressio
n
and organi
zational mod
e
l
of knowl
e
d
ge will influ
ence rea
s
o
n
i
ng efficien
cy
o
f
inferen
c
e
en
gine, affe
ct
updatin
g a
n
d
enri
c
hi
ng
knowl
edge
at
the
sam
e
t
i
me, and
aff
e
ct
intelligen
ce
le
vel of the
enti
r
e
kn
owle
dge
ba
se. A va
ri
ety of geol
ogi
cal
co
ndition
s are involved
in
line sele
ction
and desi
gn,
so there is
a large nu
m
ber of kno
w
l
edge in the route ge
ologi
cal
kno
w
le
dge
b
a
se. If all lin
e
s
-sele
c
tion
g
eologi
cal
kn
o
w
led
ge i
s
list
ed in a
knowl
edge
ba
se, it not
only increa
ses the
difficulty of mana
ging
kno
w
le
d
ge, but al
so
dire
ctly lea
d
s to d
e
cre
a
s
ed
availability and effectiveness
of se
lected lines geological
knowl
e
dge base. T
h
eref
ore, according
to its rel
a
ted
different ge
ol
ogical conditi
ons, the
kn
o
w
led
ge in th
e
kno
w
le
dge
b
a
se i
s
divid
e
d
into
a seri
es of
kno
w
le
dge
sub-spa
c
e,
su
ch
as
spe
c
i
a
l ge
otechni
cal
su
b-ba
se
, sub
-
b
a
se
of
geolo
g
ical
di
sasters. Every
kno
w
le
dge
su
b-spa
c
e
ca
n
also
be
divide
d into a
num
b
e
r of
relativel
y
indep
ende
nt kno
w
le
dge el
ements
a
c
cording
to differ
ent geol
ogi
ca
l obje
c
ts it
co
ntains. In
eve
r
y
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 7, July 201
4: 5575 – 55
84
5576
kno
w
le
dge
e
l
ement, corresp
ondi
ng el
ement-rule
s
are
save
d. Whe
n
gui
din
g
the
work of
inferen
c
e
en
gine, a
ppropriate rul
e
s will
be fo
und
in
these
kno
w
le
dge
eleme
n
ts a
s
qui
ckly
a
s
possibl
e. In a
ddition, the
knowl
edge
of
every kn
o
w
le
dge
cell i
s
stored
in the
li
ght of kno
w
le
dge
notes. T
h
is constitute
s
cla
s
s hie
r
archy
tree of lin
e
selectio
n an
d
desi
gn
kno
w
l
edge
and
th
e
sy
st
em st
ru
ct
ure
is
sho
w
n i
n
Figure 1.
Figure 1. System Structu
r
e
Diag
ram of Knowle
dge Base
As is sho
w
n i
n
Figu
re
1, knowl
edge
ba
se i
s
divid
ed
and
saved
int
o
multiple
su
b-ba
se
s
according to
the cha
r
a
c
teristics of the line se
le
ction
and de
sign g
eologi
cal kno
w
led
ge. On the
terms of
kno
w
led
ge types, it can be divided into
ge
ol
ogical kn
owl
e
dge of line se
lection field a
nd
example
kn
o
w
led
ge. Th
e
entire
kno
w
ledge
ba
se
con
s
i
s
ts of several sub
-
bases
an
d
t
h
e
kno
w
le
dge
cell co
nsi
s
ts
of kno
w
le
dg
e cell
s. Each nod
e sto
r
es 3 type
s
of kno
w
le
dg
e of
corre
s
p
ondin
g
cells.
The
s
e a
r
e
relate
d
kn
owl
edg
e
of existing
n
o
rm
s,
expe
rt experie
nces and
example
s
of existing de
sig
n
s. The
stora
ge stru
ctu
r
e o
f
knowl
edg
e cell is sh
own in Figure 2.
Unfavorable geo
l
og
y
knowledge sub
b
a
se
Route se
lec
tion
exam
ples
bas
e
Route
selec
tion
Geolog
y
Base
Inference
engine
Inter
p
rete
r
Ma
in
Module
S
p
ecia
l
T
e
rra
in
knowledge sub
b
a
se
Mollisol
Permafros
Swelling
Spe
c
i
al
ge
ot
ec
hni
c
al
knowledge sub
b
a
se
Sa
lty
Soil
Loess
……
Knowledge
Landslid
e
Mudslide
Collapse
Karst
Glaci
e
r
……
Knowledge
知
识
元
Valle
y
Mountain
Struc
t
ure
Hill
y
Reservoi
r
……
Design
Results
Explan
ation
Descri
p
tion
S
u
cces
s
f
ul
exam
ples
Failure
exam
ples
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Con
s
tru
c
tion of
Geologi
cal Knowle
dge
-b
ase
d
Syst
em
s of Rail
wa
y Route Sele
cti
on (L
v Xikui
)
5577
The e
n
tire
kn
owle
dge
ba
se con
s
ist
s
of
several
sub
-
b
a
se
s
and th
e
kn
owle
dge
cell co
nsi
s
ts
of kn
owl
edg
e
cell
s. Ea
ch
node
sto
r
e
s
3 types of
kn
owle
dge
of
correspon
ding
cell
s. T
h
e
s
e
are
related
kno
w
l
edge of existi
ng norms, ex
pert expe
rien
ce
s and exa
m
ples of exist
i
ng de
sign
s.
Figure 2. Knowled
ge Node
of Knowled
g
e
Cell
Acco
rdi
ng to
different ge
ologi
cal type
s, every
kno
w
led
ge
sub
-
base corre
s
p
ond
s to
different type
s of
ge
ologi
cal obj
ect
s
. G
e
ologi
cal
obje
c
ts in
ea
ch
category
is mad
e
a
s
a
cla
s
s
and
descri
bed
usi
ng multiple
keywords. At t
he same
time
, example
s
correspon
ding
to each categ
o
ry
are al
so in
cl
uded in its
class. A class may
concl
u
de many inst
ances a
nd a
n
example m
a
y
corre
s
p
ond t
o
multiple cla
s
ses. Th
us,
wheth
e
r be
gi
nning fro
m
cl
asse
s or fro
m
keywo
r
d
s
having
actually retri
e
ve
si
gnifica
nce, we ca
n
ea
sily
retri
e
ve kn
owle
dg
e an
d thei
r i
n
stan
ce
s.
Cl
ass’s
keyword
s
an
d
in
stan
ce
s constitute
th
e expre
ssi
on
p
a
ttern
of ea
ch type of
geol
ogical
kno
w
le
dge.
E-R rel
a
tion
ships am
ong t
hem are sh
o
w
n in Figu
re
3.
Figure 3. The
Relation
ship
among
Cla
s
ses, Keywo
r
d
s
and Instan
ce
s
Figure 4. Physical Storage St
ructure of Knowledge
Base
geological kn
owled
ge bas
e of
railway route selection
Sub-base1
……
Sub-base n
Ins
t
ance
base
Examples of existing
designs
Knowle
dge n
ode
Norm knowle
dge
Expert exper
iences
knowledg
e
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 7, July 201
4: 5575 – 55
84
5578
The mai
n
mo
dule g
eolo
g
ical kn
owl
edg
e
base of
rout
e sel
e
ctio
n consi
s
ts
of inferen
c
e
engin
e
an
d i
n
terp
reter. A
c
cordi
ng to
d
i
fferent ge
olo
g
ical
co
nditio
ns, corre
s
po
nding
kn
owl
e
dge
sub
-
ba
se i
s
i
n
voke
d. Inference Engin
e
make
s
de
cisi
on infe
ren
c
e
s
. Interprete
r i
s
respon
sible
for
explaining
reasonin
g
d
e
c
isi
on
re
sult
s a
n
d
sc
hed
u
lin
g a
p
p
r
op
r
i
a
t
e ins
t
a
n
c
e
s
an
d sho
w
s
rea
s
oni
ng de
cisi
on re
sult
s using some
form. Over
all
physical
sto
r
age
stru
cture of knowl
e
d
g
e
base is sho
w
n in Figure 4.
3. Kno
w
l
e
d
g
e
Acquisi
tion Method
Knowle
dge
a
c
qui
sition
is
a p
r
o
c
e
s
s th
at it
e
x
tr
ac
ts e
x
p
e
r
t
is
e wh
ic
h is us
e
d
to
s
o
lve
probl
em
s fro
m
the kno
w
le
dge source
s
whi
c
h ha
s thi
s
kn
owl
edg
e and convert
s
them to spe
c
i
f
ic
comp
uter
re
p
r
esentation. Geolo
g
ical
kn
owle
dge ba
se
of railway route
sele
ctio
n mainly
co
ntains
norm
kno
w
le
dge, expert knowl
edge a
n
d
instan
ce
s knowl
edge
(Sh
o
wn in Fig
u
re
5).
Figure 5. Source and
Comp
ositio
n of
Railway Route Selectio
n Geolo
g
ical Knowle
dge
Knowle
dge
a
c
qui
sition m
e
thod can be
divided into a
u
tomatic a
c
q
u
isition m
e
th
ods
and
non-autom
atic a
c
q
u
isitio
n
method
s. Aut
o
matic kn
ow
l
edge
a
c
qui
sit
i
on relate
s to
many
pro
b
le
ms
of spe
e
ch re
cog
n
ition, te
xt reco
gnitio
n
, natur
al la
ngua
ge u
nde
rstan
d
ing
an
d othe
r a
s
pe
cts.
Thus, acco
rding
to ch
aracteri
stics of
geolo
g
ic
al kno
w
le
dge b
a
se of
ro
ute
sele
ction,
t
h
e
kno
w
le
dge
a
c
qui
sition
mo
de that
com
b
ine
s
ra
w m
ode
with
ad
vance
d
m
o
d
e
is u
s
e
d
. T
he
pro
c
e
ss of
kn
owle
dge a
c
q
u
isition i
s
sh
o
w
n in Figu
re
6.
Figure 6. Pro
c
e
ss of Syste
m
Knowle
dge
Acquisitio
n
4. Kno
w
l
e
d
g
e
Repr
esen
tation of G
e
ol
ogical Rou
t
e
Selection
Knowle
dge
repre
s
e
n
tation
is
how to
sh
ow
kn
owle
dg
e (rule
s
,
con
c
ept, a
nd fa
cts) in
an
accepta
b
le fo
rm of co
mpu
t
er and info
rm peopl
e proce
s
sing
re
sults in a
way that people
can
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Con
s
tru
c
tion of
Geologi
cal Knowle
dge
-b
ase
d
Syst
em
s of Rail
wa
y Route Sele
cti
on (L
v Xikui
)
5579
unde
rsta
nd. This is the p
r
oble
m
to be studied of
knowl
edge
re
pre
s
entatio
n. It can cle
a
rl
y be
see
n
kn
owle
dge re
pre
s
e
n
tation involves two a
s
p
e
cts. They
are the form of knowl
edg
e
rep
r
e
s
entatio
n and kn
owl
edge ma
nag
ement and u
t
iliz
ation. Wh
ether the form of knowl
e
dge
rep
r
e
s
entatio
n is
rea
s
on
a
b
le or
not de
pend
s o
n
wh
ether thi
s
ex
pre
ssi
on fo
rm is
cond
uci
v
e to
kno
w
le
dge
manag
eme
n
t and u
s
e of
comp
uter.
Knowle
dge
manag
eme
n
t and utili
zation i
s
reali
z
ed by p
r
og
ram
s
, so
the form of know
l
edge
re
pre
s
entatio
n need
s to meet the form and
style of prog
ramming l
ang
uage. Ba
se
o
n
this, thi
s
p
a
ssag
e offers a ki
nd of
co
mputer i
n
tern
al
and exte
rnal
use
r
-orie
n
ted
kno
w
le
dge
repre
s
e
n
tation
form
whi
c
h i
s
suitable
for geolo
g
ical li
ne
sele
ct
ion kno
w
led
ge.
4.1. Object-o
riented
Kno
w
l
e
d
g
e Re
pr
esen
tatio
n
Inside the
Co
mputer
Curre
n
tly main kn
owl
edge
rep
r
e
s
entatio
n method
s u
s
ing m
o
re are first-ord
e
r
p
r
edi
cate
logic rep
r
e
s
e
n
tation,
pro
d
u
ction rep
r
e
s
entation,
frame repres
entation [4,
5], s
e
mantic
network
rep
r
e
s
entatio
n and
so
on.
Every metho
d
ha
s it
s a
d
vantage
s a
nd
disa
dvantag
e
s
. For
examp
l
e,
first-o
r
de
r p
r
edicate logi
c rep
r
e
s
entati
on c
ann
ot expre
ss
un
certainty kn
o
w
led
ge; fra
m
e
rep
r
e
s
entatio
n is
not g
ood at
pro
c
edural
kn
o
w
ledge re
pre
s
entation;
se
mantic network
rep
r
e
s
entatio
n exp
r
e
s
ses non
-string
e
n
c
y of
kn
owl
e
dge. Expressing
kno
w
le
dg
e with
rule
s
is
widely used
i
n
intellige
n
t engin
eeri
ng system
s,
a
n
d
alre
ady h
a
s had
a ve
ry
solid
theo
reti
cal
foundatio
n
[6, 7]. And object-o
r
iente
d
re
pre
s
entatio
n is
a mixed rep
r
esentation m
e
thod that pu
ts
together
con
v
entional re
pre
s
entatio
n
method su
ch a
s
prod
uction rep
r
e
s
entatio
n, frame
rep
r
e
s
entatio
n
and pro
c
e
ss rep
r
e
s
entation.
O
b
ject-ori
ented
rep
r
e
s
entat
ion re
prese
n
ts
kno
w
le
dge in
kno
w
led
ge b
a
se
usi
ng obj
ects
and
can
abstract
co
rresp
ondi
ng kn
owle
dge o
b
je
ct
cla
s
s for kn
owle
dge
obj
ects with
th
e same
cha
r
acte
ri
stics. It has fou
r
cha
r
a
c
teri
stics:
encap
sulatio
n
, modula
r
ity, inheritan
ce a
nd ea
sy
main
tenan
ce. Obj
e
ct-o
rie
n
ted repre
s
e
n
tation
is
clo
s
er to the human min
d
, reflects the
nature of the human thin
ki
ng pro
c
e
s
s b
e
tter and is
more
suitabl
e for knowl
edge
rep
r
esentation [8
]
For exampl
e
,
route-sele
ction geologi
cal
kno
w
led
g
e
of permaf
r
ost regio
n
s can be
expre
s
sed a
s
the following
obje
c
ts:
In the fig
u
re
above, attri
b
u
t
es of
cl
asse
s
rep
r
e
s
ent
factual
kno
w
le
dge
and
the
method
s
of cla
s
se
s
re
pre
s
ent
the
p
r
ocess knowl
edge
an
d co
ntrol kn
owle
d
ge.
Kno
w
led
g
e
cl
ass meth
ods
inclu
de the
knowl
edge fo
r
rea
s
oni
ng
su
ch a
s
all
kind
s of he
uri
s
tic
kno
w
le
dge,
meta-kno
wle
dge,
formula
s
a
nd
cha
r
t. Knowl
e
dge in
cla
s
se
s ha
s e
n
ca
p
s
ulation a
nd i
s
not allo
wed
operation o
u
t of
the obje
c
t t
o
process it
s inte
rnal
d
a
ta. Y
ou
can al
so
use
the inh
e
rita
nce
of cl
asses,
decompo
sin
g
co
mplex
knowl
edge
a
nd redu
cin
g
re
d
und
an
cy of kno
w
led
ge, to fa
cilitate
kno
w
le
dge re
aso
n
ing. Thu
s
, the static a
nd dynami
c
chara
c
te
risti
c
s of the object as a whole ex
ist
in the o
b
je
ct. The
obje
c
t
become
s
a
n
entity with kn
owle
dge
pro
c
essing
ca
pab
ilities an
d
ca
n
exist flexibly in the kno
w
le
dge ba
se.
Simultaneo
usly, it can also rep
r
e
s
e
n
t fuzzy
kno
w
le
dge to a
dd
cre
d
ibility to obje
c
t-
oriente
d
kno
w
led
ge repre
s
entatio
n. It
is shown
in
two a
s
pe
cts.
First, ea
ch
pre
c
on
dition
of
prod
uctio
n
ru
les h
a
s
different deg
ree
s
of sup
port to
con
c
lu
sio
n
s,
that is, they
have differe
n
t
Cla
ss n
a
me: perm
a
fro
s
t line sele
ction [Super class name]
spe
c
ial geote
c
hni
cal
line sel
e
ct
ion
Attributes:
Methods:
Premise list;
InputAvera
geGroun
dTe
m
perature
();
Premise wei
g
hts;
InputLineTerrain
Feat
ure
s
();
Con
c
lu
sion li
st;
InputPermafro
stZone
Type ();
Con
c
lu
sion
weights;
Calcula
t
eBeliefPropa
gationValu
e
();
Numb
er of Premise;
OutputRe
comm
end
ationsCo
nclu
sion
s()
Numb
er of Concl
u
si
on;
Outp
utCorre
s
p
ond
ingInsta
nces();
……..
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 7, July 201
4: 5575 – 55
84
5580
degree
s of importa
nce. It makes diff
erent p
r
econ
ditions have
different weights throu
g
h
assigni
ng weighting factors to preconditions. Se
cond, each
rule has diffe
rent credi
bility.
It
indicates the
degree of ce
rtainty of human expert
s
to
this rule addi
ng weig
hting factors to rule
s.
Thus it adds
the ability to
blur represented to
object-oriented knowled
ge representation
on the
basi
s
of tradit
i
onal produ
cti
on rule
s.
Rule i
s
expre
s
sed a
s
follo
w:
if
(
P
1
,
δ
p
1
)
an
d
(
P
2
,
δ
p
2
)
and
…
an
d
(
P
n
,
δ
p
n
)
then
C
1
and C
2
an
d
…
C
n
with
α
Among
this,
P
1
,
P
2
,
...
,
P
n
is prere
quisite
s;
δ
p
1
,
δ
p
2
,
...
,
δ
p
n
is right weight of
pre
r
eq
uisite
s;
C
1
,
C
2
,
...
,
C
n
is
con
c
l
u
sio
n
s of
rul
e
s;
α
is credibil
ity.
Introdu
cing
a
wei
ghting
fa
ctor of evide
n
ce
to
r
u
les
s
o
lves
no
t on
ly th
e
re
pr
es
e
n
t
a
t
ion
probl
em
s
ca
use
d
by
different
deg
re
e
s
of
impo
rta
n
ce
of m
u
ltiple evid
en
ce
s
sup
p
o
r
ting
to
con
c
lu
sio
n
s
and differe
nt indepe
nden
ce
s and
d
e
pend
en
ce
s among evid
ences, but
also
uncertain
rea
s
oni
ng p
r
obl
em with in
co
mplete evide
n
ce
s. It makes obj
ect
-
ori
ented kno
w
le
dge
rep
r
e
s
entatio
n be cap
able
of handling
fuzzy
kno
w
le
dge thro
ugh
addin
g
confid
ence sp
rea
d
i
ng
values a
nd calcul
ated con
f
idence sp
rea
d
values
to the kno
w
le
dge
obje
c
t cla
ss.
4.2. Users’ E
x
tern
al Kno
w
l
e
d
g
e Repr
esen
tatio
n
Based on Vis
u
alization
Knowle
dge in
kno
w
led
ge b
a
se, on
one
hand, ne
ed
s to be effectively store
d
, re
trieved
and id
entified
with inte
rnal
com
puter. A
t
the sam
e
ti
me, it req
u
e
s
ts to be
sh
o
w
n to
users i
n
a
more
di
re
ct
way. Mo
re
th
an 8
0
% of
h
u
man
kno
w
le
dge
and
info
rmation
is o
b
t
ained th
ro
ugh
visual, so he
re we
apply m
u
ltimedia te
chnolo
g
y to external
kno
w
le
dge
rep
r
e
s
en
tation, usin
g
not
only text message
s, but
also
3D virtu
a
l re
alit
y technolo
g
y su
ch
as
gra
phi
cs,
image
s, sou
nd,
video, anim
a
tion, etc. a
n
d
ma
king
comprehe
ns
iv
e de
scriptio
n
of multi-an
gle to
achie
v
e
visuali
z
ation
of kno
w
led
ge of
external rep
r
e
s
en
tation. External visual
repre
s
e
n
tation
of
kno
w
le
dge is
sho
w
n a
s
Fig
u
re 7.
Figure 7. External Vi
sual
Rep
r
e
s
entati
on of Knowle
dge
Figure 8. Knowled
ge Represe
n
tation Fo
rm
(a)
Text represe
n
tation of
kno
w
le
d
g
e
(b) Grap
hics
represe
n
tation
of kno
w
led
g
e
(c) Image
re
pre
s
entation
of kno
w
led
g
e
visual re
presentation of kn
owle
dge
text mes
s
a
ges
grap
hics, ima
ges
Vide
o
a
nd ani
mation
virtual reality
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Con
s
tru
c
tion of
Geologi
cal Knowle
dge
-b
ase
d
Syst
em
s of Rail
wa
y Route Sele
cti
on (L
v Xikui
)
5581
In Figu
re
7, text messag
e
s
a
r
e
used to
de
scribe
the
co
ncept a
n
d
types
of kno
w
led
ge,
and othe
r text information;
graphi
cs an
d video anim
a
tions a
r
e m
o
re ima
ge re
pre
s
entatio
n for
kno
w
le
dge. K
nowl
edge
ba
se p
r
ovid
es i
n
terfaces fo
r
input of thi
s
i
n
formatio
n. T
a
ke
kn
owl
e
d
ge
rep
r
e
s
entatio
n of valley region line
sele
cti
on for exa
m
ple and i
s
shown as Fig
u
r
e 8.
The sy
stem
build
s many types of VGE su
ch a
s
valle
y, permafro
s
t
,
loess a
nd
so on for
the pa
rticul
ar enviro
n
me
nt of ro
ute
sele
ction. It auto
m
atically invo
ke
s the
corre
s
po
ndin
g
virt
ual
geog
rap
h
ic e
n
vironm
ent from virtual
ge
ogra
phi
c
e
n
vironm
ents li
brary, expre
s
se
s
kno
w
led
ge
in
the environm
ent and then
achieve
s
virtual visual
rep
r
esentation o
f
knowl
edge.
It allows users
immersive, a
nd ma
ke
s t
hem le
arnin
g
an
d u
s
e
kno
w
le
dge
more
effecti
v
ely. The th
ree-
dimen
s
ion
a
l visual
rep
r
e
s
entation
of kn
owle
dge of
ri
ver segme
n
t’s line
sel
e
ctio
n in valley re
gion
is sh
own in
Figure 9 (Usi
ng high
brid
g
e
s a
c
ross
val
l
ey). The sy
stem can
see
dynamically 3D
kno
w
le
dge re
pre
s
entatio
n model
f
r
om d
i
fferent
pe
rs
p
e
ctives an
d d
epths an
d re
size the
virtu
a
l
geog
rap
h
ic e
n
vironm
ent.
Figure 9. 3D
Virtual Visual
Rep
r
e
s
entati
on of Knowle
dge
5. Managem
e
nt and M
a
intena
nce
In geolo
g
ical
kno
w
le
dge
b
a
se
of ro
ute
sele
ction, the
main fun
c
tio
n
s of
kno
w
le
dge b
a
se
manag
eme
n
t ha
s to b
e
done
are ba
sic man
age
m
ent of
kno
w
led
ge
(con
sist
s of a
ddi
tion,
deleting and modificatio
n
),
kno
w
le
dge demon
stratio
n
,
visual rep
r
ese
n
tation
of
kno
w
le
dge and
so o
n
. In ord
e
r to offer
users
more u
s
e
r-frie
ndl
y inte
rface
s
, the wo
rk of
kn
owle
d
ge man
age
m
ent
se
ction is
mainly focu
sed on reali
z
ing vis
ual repre
s
e
n
tation
of knowle
d
ge and sim
p
le
decompo
sitio
n
of general knowl
edge in
p
u
t by users.
Figure 10. Input Interface of
Knowle
dge
Figure 11. Input Interface of
Examples
(b)
3D virtual
rep
r
e
s
entatio
n
of Knowled
g
e
(Vertigo
Mode
)
(a)
3D virtual rep
r
esentation
of Knowled
g
e
(Gi
z
mos Mode)
(c) 3D virtual
rep
r
e
s
entatio
n
of Knowled
g
e
(Texture
Mode
)
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 7, July 201
4: 5575 – 55
84
5582
The inte
rfa
c
e
that users i
nput
kno
w
le
dge i
s
sho
w
n a
s
Fig
u
re
10. When i
n
i
t
ializing
kno
w
le
dge b
a
se, a
n
expe
rt or a kno
w
le
dge en
gine
er
enters a rule
usin
g the u
s
er inte
rface, and
this rul
e
is ju
dged by the
system. If the rule’
s
p
r
em
ise o
r
co
ncl
u
sion
contai
ns both "And" and
"Or", the system will automatically convert it into
a
multi-decom
p
osition ru
le and store into the
kno
w
le
dge b
a
se to fa
cilita
t
e future mai
n
tenan
ce of
knowl
edge
ba
se. The
rule
s stora
ge thro
ugh
this form
is conve
n
ient
for rul
e
s re
d
unda
nc
y
che
cki
ng, contra
dictory
rule
s che
c
king
a
nd
circulatio
n rul
e
s che
cki
ng.
To achieve fuzzy kno
w
le
d
ge re
pre
s
e
n
tation, every pre
r
eq
uisite
of a
rule h
a
s o
w
n
cre
d
ibility input value an
d the maxi
m
u
m value is
1. After sele
cting a pi
ece
of
kno
w
le
dge, we can add
instan
ce
s to it through
fu
nction ‘Add
an insta
n
ce’. This process is
sho
w
n in Fig
u
re 11.
6. Reaso
nin
g
of Kno
w
l
e
dge Ba
se
6.1, Reaso
ning of Un
cer
taint
y
In enginee
rin
g
probl
ems, t
here i
s
much
commo
n kn
owle
dge which can
not use
normal
logic to ha
ndl
e, beca
u
se they contai
n a great d
eal
of unce
r
tainty
[9].
For example, observe
the
followin
g
pie
c
es of kno
w
le
dge: ‘Rel
atively flat
side of the river b
ank m
a
y have bad g
eolog
ical
con
d
ition
s
’, ‘The terrain of valley side
s is
steep
a
nd it may cau
s
e g
eologi
cal di
sa
sters land
slid
es,
mudsli
de
s, et
c.’. The
s
e pi
e
c
e
s
of
kno
w
l
edge
co
nt
ain
a great de
al
of uncertainty
. There
are two
cla
s
ses of th
e uncertainty
above. The
y
are uncert
a
inty of rule con
d
ition
s
an
d uncertainty
o
f
con
c
lu
sio
n
s.
Therefore, it
is n
e
cessa
r
y to bu
ild
so
me un
ce
rtain
pro
c
e
s
s of
cal
c
ulatio
n a
nd
rea
s
oni
ng [10
-
12].
1) Un
ce
rtaint
y of rule cond
itions
Whe
n
o
b
serving obje
c
t
s
, the truth
that we
h
a
ve seen
u
s
u
a
lly owns
u
n
ce
rtainty.
Gene
rally, un
certai
nty of a truth is d
e
scribed
by
a co
efficient ra
ngi
ng from 0 to
1.’1’ rep
r
e
s
en
ts
compl
e
te det
ermin
a
tion a
nd ‘0’ repre
s
ent
s compl
e
te un
certai
n
t
y. This coef
ficient is
call
ed
cre
d
ibility. When
a rule
h
a
s
more tha
n
one
conditi
o
n
, you n
eed
to cal
c
ul
ate
credibility of t
h
e
overall condit
i
on acco
rding
to credibility of each
credi
bility[13-14]. There are two main ways
as
follow:
a) The a
ppro
a
ch b
a
sed on
fuzzy set the
o
ry
Accordi
ng to this
appro
ach, it regards
the sm
allest credibility
of all
the condit
i
ons
as
cre
d
ibility of t
he ove
r
all
co
ndition. Fo
r i
n
stan
ce,
thi
s
approa
ch i
s
u
s
ed
by MY
CIN
system. Se
t a
rule containi
ng m con
d
itions, and
m
c
c
c
,...,
,
2
1
is each condition’s credi
b
ility. Therefore,
credibility of the overall condition
t
c
is
}
,...,
,
min{
2
1
m
t
c
c
c
c
. As is sh
own in
Figure 12, th
ere
are
3 rules owning rules.
Assuming
that 0.9, 0.5 and
1.0
are credibility of the th
ree rules, so take
the minimum
value 0.5 for the credibility of the total.
Figure 12.
Fu
zzy
-set
Proce
ssi
ng Metho
d
of Condition
Cre
d
ibility
b) The approach ba
sed on probability
This approach also gives
all evidences its
own credibility. But credibility of the overall
con
d
ition equ
als to the total prod
uct of every cr
edibil
i
ty. For example, this app
roach is used
b
y
PROSPECTOR s
y
s
t
em. Us
ing
the s
a
me
rule as
is
s
h
own in Figure
12, the overall
c
r
edibility of
part of the rul
e
s ente
r
ing i
s
0.45, as is shown in Figu
re 13.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Con
s
tru
c
tion of
Geologi
cal Knowle
dge
-b
ase
d
Syst
em
s of Rail
wa
y Route Sele
cti
on (L
v Xikui
)
5583
Figure 13.
Probability Pro
c
essing Meth
o
d
of Conditio
n
Cre
d
ibility
For the same
rule having
m conditio
n
s,
m
c
c
c
,...,
,
2
1
is each
condition’s
credibility. So
credibility of the overall
condition
t
c
is
m
i
i
t
c
c
1
.
Therefore, thi
s
arti
cle prov
ides two methods to determ
i
ne
the overal
l credi
bility of
the
rule
s se
ction;
you can cho
o
se o
ne a
c
co
rding to the a
c
tual situ
ation
.
6.2. Uncer
tai
n
t
y
on Conclusions
The u
n
certai
nty on
con
c
l
u
sio
n
s is
also call
ed
un
certainty of
ru
les, a
nd it
re
pre
s
ent
s
uncertain de
gree of
pro
d
u
cin
g
so
me con
c
lu
sio
n
when the
con
d
itions of th
e rule a
r
e f
u
lly
compl
e
ted. It is also
rep
r
ese
n
ted by given ru
le
s
coefficient
s be
tween 0 a
n
d
1, that is, the
coeffici
ent is
cre
d
ibility of a con
c
lu
sion.
Determinatio
n method a
s
follows:
Take
con
c
lu
sion credibility
as the pro
d
u
ct of
con
d
itions
cre
d
ibilit
y and rule
s coefficien
t
in
out
C
C
. As is
sho
w
n in Fig
u
re
1
4
, at this tim
e
co
ndition
s
cre
d
ibility is
0.5 and
rul
e
s
coeffici
ent is
0.8, so
co
ncl
u
sio
n
cre
d
ibil
ity is
4
.
0
8
.
0
5
.
0
out
C
. The rel
a
tionship b
e
twee
n
conditions credibility of rules a
nd
conclusion credibi
lity of rule
s i
s
shown i
n
Figure 14.
T
h
is
relation
shi
p
can be u
s
ed to
represent un
certai
nty of rules.
Figure 14. Ca
lculatio
n method of Co
ncl
u
sio
n
Credibility
Figure 15. Th
e Relatio
n
ship betwe
en
Conditions Credibility of
Rules and Concl
u
sion
Credibility of Rules
7. Conclusio
n
Acco
rdi
ng to
the diversity of geol
ogica
l con
d
it
ion
s
i
n
volved in
ra
ilway lo
catio
n
de
sign,
different types of geologi
cal object
s
we
re cl
a
s
sified
and indexe
d
, and pro
pose
d
the modeli
ng
method ba
sed on multiple sub
-
b
a
ses. It used
the compu
t
er's inte
rnal
and external
rep
r
e
s
entatio
n
in
ge
ologi
cal
kno
w
led
g
e
re
pre
s
e
n
tation. In comp
uter external,
it used obje
c
t-
oriente
d
cla
s
s-rule
s kno
w
l
edge
rep
r
e
s
e
n
tation mod
e
l
, added
cre
d
ibility to the object
-
ori
ent
ed
kno
w
le
dge
re
pre
s
entatio
n
model, an
d a
l
so can re
p
r
e
s
ent fu
zzy re
pre
s
entatio
n
of kno
w
led
g
e.
Achieve visu
al rep
r
esenta
t
ion of external kno
w
led
g
e
representati
on usin
g text
and graphi
cs 3D
virtual re
ality tech
nology.
According t
o
the a
m
big
u
ity of geolo
g
ical
kn
owl
e
dge, it reali
z
es
uncertainty reasonin
g
of the part b
a
se
d on
rul
e
co
ndition and con
c
lu
si
ons and rea
lize
s
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NI
KA
Vol. 12, No. 7, July 201
4: 5575 – 55
84
5584
rea
s
oni
ng
ba
sed
on
in
stan
ce
s. G
eologi
cal kno
w
le
dg
e-
b
a
s
e
d s
y
s
t
ems
o
f
ra
ilway ro
ute
sele
cti
on
build
s a full range of
rout
e-sele
ction g
eologi
ca
l e
n
vironm
ents fo
r route
sele
cti
on and
de
sig
n
,
provide
s
real
-time hel
p a
nd refere
nce
for e
ngine
ers a
nd Better meets the
requireme
nts
of
railway geolo
g
ical route se
lection.
Ackn
o
w
l
e
dg
ements
This research wa
s sup
p
o
r
ted by the Nati
on
al Nat
u
ral Scie
nce
Foundatio
n of China
(Proje
ct No.:
5127
8316
), the People’
s
Rep
ubli
c
of China.
Referen
ces
[1
]
ZH
AO Ji
a
n
c
han
g
,
L
I
U
Shi
z
hon
g
,
JI Su
i
w
ang. En
g
i
ne
e
r
in
g
g
e
o
l
o
gi
ca
l p
r
ob
l
e
ms in
rai
l
w
a
y
en
gi
n
e
e
r
in
g
construction.
T
he Ch
ines
e Jo
urna
l of Geolo
g
ical H
a
z
a
r
d
a
nd Co
ntrol.
20
01; 12(1): 7-9.
[2]
YI Sirong, Z
H
UANG Haiz
he
n, HAN Ch
un
h
ua. An
E
x
p
e
rt-H
yperte
xt Mo
d
e
l for Kn
o
w
l
e
d
ge Obtai
nac
e
and Ali
gnm
ent
Guidanc
e in Rail
w
a
y Loc
ati
on S
y
stem.
R
a
ilw
ay Investig
ation a
nd Sur
v
eyin
g
. 2005
;
31(1): 1-4.
[3]
SHI W
enhui.
W
o
rk on the ra
il
w
a
y e
n
g
i
ne
eri
ng ge
ol
og
y an
d road
be
d.
Jou
r
nal of Ra
ilw
ay
Engin
eeri
ng.
198
8; 3: 7-12.
[4]
W
A
NG Kehon
g, T
A
NG Z
h
izhong, H
U
Pe
ng,
etc.
Kno
w
l
e
d
g
e
en
gi
neer
ing
and k
n
o
w
l
e
dg
e proc
essin
g
s
y
stem. BEIJING:
T
s
inghua
Univers
i
t
y
Pr
es
s. 1994.
[5]
W
U
Z
hao, XIA
N
G Jingme
ng.
An obj
ect-ori
e
n
t
ed fuzz
y
k
n
o
w
led
ge b
a
se mo
del.
Co
mputer app
licati
o
n
s
and softw
are.
200
5; 22(1): 22
-24.
[6]
Roston
D. Pri
n
ciples
of Artific
i
al Int
e
lli
ge
nc
e a
n
d
Ex
pe
rt Sy
ste
m
D
e
vel
opme
n
t. N
e
w
York: McGraw
-
Hill. 19
88.
[7]
JIA Yan, W
A
NG Z
h
i
y
ing. T
he pr
inci
ple
a
nd
tech
nol
og
y of kno
w
l
e
d
g
e
base s
y
stem,
Cha
ngsh
a
:
Natio
nal U
n
ive
r
sit
y
of Defe
ns
e T
e
chnolo
g
y
Press. 1993.
[8]
F
A
N Yuehu
a, Z
H
ANG Kaibin
g. Object orien
t
ed
kno
w
l
e
d
ge
repres
entati
on of W
eapon F
a
ult Diag
nos
i
s
Ex
pert S
y
stem.
Journa
l of Xi'
an T
e
chno
lo
gica
l Univers
i
ty
. 20
08; 22(2): 1
68-
171.
[9]
YI Sirong. Pre
d
icate l
ogic kn
o
w
le
dg
e e
x
pr
e
ssion mo
del of
object-ori
ente
d
class an
d its appl
icatio
n i
n
Lin
e
Selecti
on
S
y
stem.
Chi
na Railw
ay
Sci
enc
e.
2003; 2
4
(6): 31-3
5
.
[10]
CAI Z
i
xin
g
, Du
rkin J, GONG
T
ao. Senior
e
x
p
e
rt
s
y
stem:
princi
pl
e, d
e
si
gn
an
d a
p
p
lic
ation. B
e
ij
in
g:
Scienc
e Press. 2005; (8).
[11]
Shufan
g Z
H
ANG, Jun HAN. Bu
ildi
ng Structured Pe
er-to-Peer R
e
sour
ce Shari
ng Pl
atform Us
i
n
g
Object Enca
ps
ulati
on Ap
pro
a
c
h.
T
E
LKOMNIKA Indones
ia
n Journ
a
l
of
Electrical E
n
g
i
n
eeri
n
g
. 20
13;
11(2): 93
5-9
4
0
.
[12]
Xu
e
x
i
a
Li
u. Stud
y
on K
n
o
w
le
dg
e-b
a
sed
In
telli
gent F
ault Di
ag
nosi
s
of H
y
drau
l
i
c S
y
stem
.
T
E
LKOMNIKA Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
ng.
2012; 1
0
(8): 2
041-
204
6.
[13]
I ketut Gede D
a
rma Putra, Pu
tu Manik Pri
hat
ini.
F
u
zz
y E
x
p
e
rt S
y
stem for
T
r
opical Infecti
ous Dis
eas
e
by
C
e
rtai
n
t
y
F
a
cto
r
.
T
E
LKOMNIKA Indo
ne
sian J
our
nal
of
Electrica
l
En
gi
neer
ing.
20
12;
10(6):
152
6-
153
6.
[14]
Abdu
lla
h Al
ali.
Automatic S
egme
n
tatio
n
F
r
am
e
w
ork
of Buil
din
g
An
a
t
omical M
ous
e Mod
e
l fo
r
Biol
umin
escen
c
e T
o
mograp
h
y
.
T
E
LKOMNI
KA Indo
nesi
a
n
Journ
a
l
of
El
ectrical E
n
g
i
ne
erin
g.
20
13
;
11(9): 49
97-
50
04.
Evaluation Warning : The document was created with Spire.PDF for Python.