TELKOM
NIKA
, Vol.13, No
.3, Septembe
r 2015, pp. 7
76~782
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v13i3.1804
776
Re
cei
v
ed Ap
ril 25, 2015; Revi
sed
Jun
e
21, 2015; Accepted July 1
3
,
2015
Failure Mode and Effect Analysis of Power Transformer
Based on Cloud Model of Weight
Jianpeng
Bian*, Xiao
y
u
n
Sun, Jing Yang
Schoo
l of Elect
r
ical a
nd Electr
onics En
gin
eer
ing, Shi
jiaz
h
u
a
ng T
i
edao U
n
iv
ersit
y
,
Shiji
azh
u
a
ng, 050
04
3, Hebe
i, Chin
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: bjp2
10@
qq.c
o
m
A
b
st
r
a
ct
As the k
e
y com
p
onent of
a
power system
,
the power transforme
r direct
ly im
pacts the
reliability
and safety of the syste
m
. Failure
mo
de a
n
d
effects anal
ys
is (FMEA) is a meth
odo
lo
gy used to a
naly
z
e
potential failur
e
m
o
des within
a syst
em and has been
used
extensiv
ely to
exam
ine the power transfor
m
er
’
s
perfor
m
a
n
ce i
n
vario
u
s pote
n
tial fai
l
ure sc
enar
ios.
How
e
ver, the F
M
EA meth
od h
a
s severa
l flaw
s; for
exa
m
p
l
e, the n
on-d
i
fferenti
a
l
ana
lysis
of eva
l
uati
on in
dex a
nd the i
m
p
o
ssi
b
ility of eva
l
u
a
ting the act
ual r
i
sk
amon
g risk pri
o
rity nu
mb
er (
R
PN) val
ues t
hat on
th
e su
rface are
equ
al. T
he clo
ud
mo
de
l of w
e
ig
ht
incor
porates th
e relative i
m
po
rtance of ind
e
x
. T
h
is paper prop
oses ap
pl
ying F
M
EA ba
sed on the cl
o
u
d
mo
de
l of w
e
ig
ht to ass
e
ss
a p
o
w
e
r trans
former
for
r
i
sk, and
sh
ow
s that this
meth
o
d
ca
n effectiv
el
y
overco
me the defects of tradi
tiona
l F
M
EA assessment meth
ods.
Ke
y
w
ords
: Po
w
e
r transforme
r
, Maintena
nce
strategy, Clou
d
mo
de
l of w
e
ight, F
M
EA
Copy
right
©
2015 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
The p
o
wer t
r
ansfo
rme
r
i
s
the key equi
p
m
ent in
a po
wer sy
stem; t
herefo
r
e, th
e
norm
a
l
operation of t
he tran
sfo
r
m
e
r is critical f
o
r t
he
safety
and
stability of the syste
m
. Failure
m
ode
and effe
cts
analysi
s
, first
develop
ed
at Grum
man
Aircraft Co
rporatio
n in t
he 19
50
s, is the
methodol
ogy
most com
m
only use
d
to perform pre
v
entive maintenan
ce. Thi
s
mainten
a
n
c
e
identifies an
d
eliminate
s
kno
w
n and/o
r
pote
n
tial
fa
ilure
s, be
gin
n
ing by
ran
k
ing the
high
est-
prio
rity issue
s
[1, 2]. Trad
itional FMEA
determi
ne
s t
he ri
sk p
r
iorit
i
es of fail
ure
mode
s by u
s
i
n
g
the ri
sk
prio
ri
ty number,
which i
s
d
e
termined by
ea
ch ri
sk fa
cto
r
’s occu
rre
nce (O), severity (S)
and dete
c
tion
(D). Tradition
al RPN is a p
r
odu
ct
of these three fa
ctors [3, 4]. That is:
RPN
O
S
D
(1)
The d
r
a
w
ba
cks of u
s
in
g
RPN to
prioriti
ze fail
ure mode
mai
n
tenan
ce fo
r power
transfo
rme
r
parts are cle
a
r
[5-8].
Th
e
y
include
pri
m
arily differe
nt sets of ri
sk fa
ctors m
a
y
prod
uce the
same
RPN val
ue, but thei
r
risk impli
c
atio
ns m
a
y be
qu
ite different.
And the
relati
ve
importa
nce of O, S and D is not take
n in
to account.
The fuzzy set theory, whi
c
h ha
s
demonstrated great
capability and perf
o
rm
s i
n
a variety
of applicatio
n
domain
s
su
ch as control
and mod
e
ling
[9, 10], can captu
r
e the u
n
ce
rtainty an
d
ambiguity of factors. Th
us,
the fuzzy RP
N ha
s
b
een
widely utilize
d
in FMEA to
overcome
so
me
of tradition
al
FMEA’s afo
r
e
m
entione
d d
r
awb
a
cks
[11, 12].
Ho
weve
r,
app
roa
c
he
s based prim
ari
l
y
on proba
bility or fuzzy set theory u
s
ually igno
re
all un
certai
nties that m
a
y occur
du
ring
the
evaluation
proce
s
s. In ad
dition,
the fu
zzy
RPN m
odel la
cks a
n
effective
way to tran
slate
qualitative evaluation to q
uant
itative numeri
c
al val
ue. Thus, F
M
EA based
on fuzzy the
o
ry
can
not atta
ch en
oug
h i
m
porta
nce t
o
un
ce
rtainty
to a
dequ
ately asse
ss
and
prio
ritize
ri
sk
maintena
nce in the power tran
sform
e
r.
This pa
pe
r propo
se
s a novel FMEA method, ba
sed o
n
the cloud m
odel of weig
h
t, which
will fully recogni
ze
and i
n
co
rpo
r
ate
the imp
o
rta
n
ce of u
n
certa
i
nty in the
ri
sk a
s
sessme
nt
pro
c
e
ss. Thi
s
method overcome
s the d
e
fects of
trad
itional FMEA assessme
nt methods (no
n
-
differential
a
nalysi
s
of
evaluation
inde
x and th
e d
a
nger of
assu
ming that
eq
ual
RPN value
s
indicate equal amounts of
actual
ri
sk)
and thereby
i
m
proves the
credibility of
risk assessm
e
nt.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Failure Mode
and Effect Analysis of Power Tran
sform
e
r Based on
Clou
d… (Jian
peng Bian
)
777
The p
r
op
ose
d
metho
d
a
l
so p
r
ovide
s
a c
onvinci
ng foun
datio
n for pl
anni
ng an
effici
ent
maintena
nce strategy, which will improve
the transfo
rmer’
s
se
cu
rity and value.
The
re
st of th
is p
ape
r i
s
o
r
gani
zed
as fo
llows. Sectio
n 2 d
e
scribe
s
the FMEA b
a
se
d on
the clou
d mo
del of wei
ght method. Se
ction 3 illu
st
rat
e
s the p
r
a
c
tical appli
c
ation
of this meth
od
to assess the
powe
r
tran
sf
orme
r. Sectio
n 4 pre
s
ent
s con
c
lu
sio
n
s.
2. FMEA Ba
s
e
d on Cloud
Model of We
ight
2.1. Ev
aluation Index of
Po
w
e
r Tran
s
f
ormer
In orde
r to properly a
s
sess the po
wer t
r
an
sf
orm
e
r,
we mu
st esta
blish the
key index of
the asse
ssm
ent. It would
not be p
r
a
c
tical to in
clu
d
e
every po
ssi
b
l
e risk fa
ctor
in this exam
p
l
e.
Instead,
key index we
re
ch
ose
n
that are
broa
d
en
oug
h to encomp
a
ss th
e ent
ire
stru
cture of the
power tra
n
sfo
r
mer, rep
r
ese
n
t each
critical par
t and p
r
odu
ce a com
p
reh
e
n
s
ive ri
sk a
s
se
ssme
nt.
There are
m
any possible failure modes i
n
a transform
e
r,
and their probability of
occurre
n
ce, d
e
tection
an
d
severity
differ. Base
d
o
n
statistical
data
and expert
op
inion,
traditio
nal
FMEA cal
c
ul
ates th
e p
r
o
bability of e
a
ch
failure
mode
and
ranks th
ese
prob
abilitie
s i
n
to 5
grad
es.
The
highe
r the
g
r
ade i
s
, the
g
r
eate
r
the
po
tential da
nge
r to th
e p
o
wer tran
sform
e
r
(Tabl
e 1-3
)
.
Table 1. Evaluation criteri
a
of severity
Sever
i
ty
Effect of sever
i
ty
Ranking
Hazardous
Tremen
dous loss of environment
and personnel, e
x
treme
damage t
o
transforme
r
,
w
ill
gr
eatl
y
affect pow
e
r
s
y
stem
10
Ver
y
severe
high failure rate
of the po
w
e
r tra
n
s
former and m
a
y cause loss of environment and
personnel
8,9
Comparativel
y
sever
e
Significant damage to transfo
rmer
, unc
lear to
w
hat
degree environ
m
ent and
personnel w
ill
suffer
6,7
Moderate
Some damage to
transforme
r
, likely some loss of environment and
personnel
2,3,4,5
Lo
w
Transform
er an
d
po
w
e
r s
y
stem
no
t affected, no obv
ious damage to e
n
vironment
and personnel
0,1
Table 2. Evaluation criteri
a
of detection
Detection
Likelihood of detection
Ranking
Absolute
uncertaint
y
Potential failure or fault ma
y be v
e
r
y
difficult to det
ect and (once de
tected) ma
y
require eme
r
gen
cy
m
a
jor ref
u
rbishment
10
Ver
y
lo
w
Potential failure or fault ma
y be d
e
tected
w
i
th strict online multi-monitoring
8,9
Moderate
Potential failure ma
y
be d
e
tected
w
i
th strict overall monitoring and in
creasing the
number of diag
n
o
stics performed
5,6,7
High
Potential failure ma
y
be e
a
sily
de
tected
w
i
th strict monitoring schedule
2,3,4
Almost certain
Potential failure ma
y
be e
a
sily
de
tected b
y
the
app
earance, sound
and temper
ature
of transform
0,1
Table 3. Evaluation criteri
a
of occu
rren
ce
O
ccurrence
Probabilit
y
of occurrence
Ranking
Extrao
rdinaril
y
hi
gh
Often
10
High 0.5~1
times/
y
ear
8,9
Moderate
Periodically
5,6,7
Lo
w 1~3
times/year
2,3,4
Ver
y
lo
w
3~5 times/year
0,1
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 776 – 782
778
2.2. Cons
tru
c
tion of the
Cloud Mod
e
l of Weigh
t
Clou
d model
is a ne
w m
odel for
con
c
ept
rep
r
e
s
ent
ation and
ca
pture
s
the u
n
ce
rtain
transitio
n be
tween qu
alitative conce
p
t and it
s quantitative rep
r
e
s
entatio
n. The digital
cha
r
a
c
teri
stics of cl
oud
s
can integ
r
ate
the fu
zzine
ss and rand
om
ness of lin
gu
istic te
rms i
n
a
unified way, whi
c
h
l
a
ys a foundatio
n
of
kn
owl
edg
e
repre
s
e
n
tation
. Therefore, the
clou
d m
o
del
can mi
mic a
human
bein
g
’
s thin
king a
n
d
is mo
re
fle
x
ible and eff
e
ctive than t
he conventio
nal
fuzzy rea
s
o
n
ing meth
od
s in rep
r
e
s
enting u
n
ce
rtainty and
prop
agatin
g
kno
w
le
dge.
The
cal
c
ulatio
n method of the cloud mod
e
l of weight
is b
a
sed on the pro
duct of the sq
uare
root. Th
e
s
p
ec
ific
pr
oc
es
s
is
a
s
fo
llow
s
[1
3
]
:
(1) Con
s
tru
c
tion
of the cloud
matrix of cont
rast
Every elem
e
n
t in th
e
clo
ud m
a
trix of
contrast
is constituted
by
the
clo
ud m
odel. T
h
e
element
ik
C
rep
r
ese
n
ts the
contributin
g d
egre
e
of the sub
-
ind
e
x
i
f
to the prio
r-i
nde
x, which i
s
relative to
k
f
. Assume the n
u
mbe
r
of the sub-i
ndex is
n
. Throu
gh ana
lyzing the ch
ara
c
teri
sti
cs
of the pai
r-wi
se
com
p
a
r
iso
n
cl
oud
matri
x
, it is cle
a
r t
hat the di
ago
nal ele
m
ent
ii
C
re
pre
s
ent
s th
e
importa
nt de
gree
of com
pari
s
on, that
is,
1
ii
Ex
,
0
ii
En
,
0
ii
He
. Again, the elem
e
n
t
ik
C
rep
r
e
s
ent
s the contri
buting
degre
e
of the sub
-
ind
e
x
i
f
to the prio
r-in
dex, which is relative to
k
f
and the
sym
m
etrical elem
ent
ki
C
is
the
c
ontras
t. So, the
ik
C
and
ki
C
mus
t
satis
f
y the following
relation:
11
1
1
1
1
1
1
1
1
1
1
1
1
1
1n
1
1
1
1
,,
C
(
,
,
)
,
,
(
,
,
)
=
,,
C
(
,
,
)
,
,
(
,
,
)
nn
n
n
n
n
n
n
n
n
n
nn
nn
nn
nn
C
C
Ex
En
He
C
E
x
E
n
H
e
C
C
C
Ex
En
He
C
E
x
E
n
H
e
(2)
22
11
(,
,
)
()
()
ki
ik
ik
ik
ik
En
He
cC
cE
x
E
x
E
x
(3)
(2) Valu
e of the eleme
n
ts i
n
the clou
d matrix of contra
st
Whe
n
con
s
tructing
the
clo
ud m
a
trix of
cont
rast, th
e
mutually imp
o
rtant
deg
ree
s
of
th
e
different in
de
x shoul
d b
e
determi
ned fi
rst to fo
rm t
he cl
oud
mo
del, that is, t
he cl
oud’
s
p
o
le.
Secon
d
ly, experts mu
st d
e
termin
e the
deg
ree
of
e
a
ch
sub-i
nde
x’s impo
rtan
ce rel
a
tive to
the
clou
d’s p
o
le. Finally, the re
sults a
r
e
synthesi
z
e
d
.
The
con
c
rete process for a cl
ou
d weig
ht based
FMEA analysis of a power
transfo
rme
r
is as follows:
The d
egree o
f
mutual imp
o
r
tance of FM
EA’s th
re
e ri
sk facto
r
s (S,
O, and
D) is
divided
into five ran
k
s. Experts d
e
t
ermine the
mutually
imp
o
rtant de
gre
e
of each fa
ct
or relative to the
clou
d model’
s
pole. The qu
alitative result
s are
q
uantifi
ed throu
gh th
e application
of synthesi
z
e
d
,
multi-expe
rt
quantification
to dete
r
mi
n
e
the valu
e of
ea
ch fa
ctor.
The
Ex
and
of the cl
oud
model of different impo
rtan
t degree i
s
listed in Table 4
.
And
.
Table 4. Division of importa
nt degre
e
an
d corre
s
p
ondi
ng expe
ct value and e
n
tro
p
y value of
clou
d model
Degree o
f
importance
Same A
little
important
Obviously
important
Intensel
y
important
Extremel
y
important
1
1.6240
2.1204
4.3528
6.4218
En
1.4587
2.3603
3.8194
6.1820
9.8556
(3)
Cal
c
ulatio
n of cloud of
weig
ht
The
clo
ud m
odel
of weigh
t
can
be
com
p
reh
end
ed
a
s
m
odul
ating
the tra
d
itional
wei
ght.
In the chara
c
teri
stic num
bers of the cloud mod
e
ls
(
,
En
,
He
), the
i
s the weight of the
traditional
FM
EA analysi
s
,
and th
e
En
and
are utilized to fine-tune the value
of the
weight by
adju
s
ting the
para
m
eters. The co
ncrete
pro
c
e
ss i
s
as
follows:
(1)
Cal
c
ulate
the prod
uct o
f
every element of
the clo
ud matrix of contrast. Assume the
prod
uct of all element
s of row
i
is
(,
,
)
ii
i
i
M
Ex
En
H
e
, that is
:
En
0.00
5
He
Ex
Ex
Ex
He
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Failure Mode
and Effect Analysis of Power Tran
sform
e
r Based on
Clou
d… (Jian
peng Bian
)
779
0
(,
,
)
jn
ii
i
i
i
j
j
M
Ex
En
He
C
(4)
The definitio
n of the multip
lication
of the clou
d model
11
1
1
(,
,
)
CE
x
E
n
H
e
and
22
2
2
(,
,
)
CE
x
E
n
H
e
is:
12
22
12
12
1
2
12
22
12
12
12
()
(
)
()
(
)
x
n
e
Ex
Ex
E
En
E
n
CC
E
E
x
E
x
Ex
E
x
H
He
H
e
Ex
Ex
Ex
Ex
(5)
(2) Cal
c
ulate the
n
th root of
(6)
(3)
Normali
z
a
t
ion of vector
W
The no
rmali
z
ation of vect
or
W
is
12
(,
,
,
)
n
WW
W
W
.
i
W
is the cl
oud mo
del of
weight
i
index. And
i
W
sa
t
i
sf
ies
1
i
i
n
i
i
W
W
W
(7)
The definitio
n of the ad
d
operation a
nd t
he divi
si
on op
eratio
n
of the clo
u
d
model
11
1
1
(,
,
)
CE
x
E
n
H
e
and
22
2
2
(,
,
)
CE
x
E
n
H
e
is:
12
22
12
1
2
22
11
x
n
e
Ex
Ex
E
CC
E
E
n
E
n
H
He
He
(8)
12
22
11
2
21
2
22
12
12
/
()
(
)
()
(
)
x
n
e
Ex
Ex
E
CE
n
E
n
E
CE
x
E
x
H
He
H
e
Ex
Ex
(9)
2.3. Process
of FMEA
Ba
sed on Clou
d Model of
Weigh
t
(1) A
cco
rdi
n
g
to Table 1-3,
the severity, detectio
n
and
occurren
ce
of potential failure in
comp
one
nts
of the p
o
we
r tra
n
sfo
r
me
r are
re
spe
c
tively evaluat
ed a
nd th
e
co
rrespon
di
ng
quantitative value
s
are d
e
termin
ed.
(2) T
he clo
u
d
weight of sev
e
rity, detectio
n
and o
c
curre
n
ce
can b
e
g
o
t base
d
on (1)-(5
).
(a) A
c
cordin
g to Tabl
e
4, the intert
wine
d impo
rt
ance of severity, detecti
on an
d
occurre
n
ce
a
r
e
jud
ged by experts, and the
value
of
each elem
ent
acco
rding t
o
the clo
ud m
a
trix
of contrast
C
is determi
ned.
11
/
1
1
/
(,
,
)
=
(
,
,
)
n
ii
i
i
i
ii
n
i
nn
ii
WM
E
x
E
n
H
e
En
He
Ex
nE
x
n
E
x
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 776 – 782
780
(b) Th
e prod
uct
(,
,
)
ii
i
i
M
Ex
En
H
e
of every element of the
clou
d matri
x
of contrast
C
is
cal
c
ulate
d
through Equ
a
tio
n
(4)-(5).
(c) T
he nth
root
12
3
(,
,
)
WW
W
W
of
ca
n b
e
calculated
according
to t
he
Equation (6).
(d) Th
e
i
s no
rmali
z
ed by
Equation (7)-(9) to produ
ce the qu
an
titative value o
f
severity, dete
c
tion an
d occurren
ce in the
cloud mat
r
ix of contrast.
(e)
The
qua
ntified value
s
o
f
severity, det
ection
and
occurre
n
ce a
c
cordin
g to
step
(1) are
norm
a
lized a
nd ma
ppe
d o
n
to the
clo
u
d
wei
ght of
se
verity, detecti
on a
nd
occu
rren
ce. T
hen
the
weig
ht of severity, detectio
n
and o
c
curre
n
ce
can b
e
calcul
ated.
(3) Acco
rdin
g
to the Equation (10
)
, RPN
c
an be
cal
c
ul
ated throu
gh
the quantified
values
for severity,
d
e
tection, occurren
ce deriv
ed
a
c
co
rdin
g
to step
(1) an
d the
clo
ud
weight of
seve
rity,
detectio
n
and
occurren
ce.
so
d
RP
N
w
S
w
O
w
D
(10)
(4) A
c
cordi
n
g
to Table 5 a
nd RP
N, the corre
s
p
ondin
g
mainten
a
n
c
e strate
gy for each
part of the po
wer tran
sform
e
r ca
n be det
ermin
ed.
Table 5. Co
rresp
ondi
ng M
a
intena
nce Strat
egy acco
rd
ing to the thre
shol
d of RPN
RPN<2
2<RPN<4
4<RPN<6
6<RPN<8
8<RPN<10
Maintenance
str
a
tegy
Correc
t
i
v
e
Maintenance
Extended
Maintenance
Time-based
maintenance
Maintain as quic
k
ly
as possible
Maintain
Immediatel
y
3. Practical
Applica
t
ions
of FMEA Ba
sed on Clou
d Weight
Let us
co
nsi
d
er a
sam
p
le
500
kV po
we
r transfo
rme
r
and a
nalyze i
t
using F
M
EA base
d
on the
clo
u
d
weig
ht mod
e
l. The
ope
rational hi
sto
r
y and m
a
inte
nan
ce
re
cord
s of o
u
r
sa
m
p
le
transfo
rme
r
show that the transfo
rme
r
h
a
s ex
pe
rien
ced a sho
r
t-te
rm eme
r
ge
ncy load and h
a
s
had on
e overhaul. The ma
jor problem
with this tra
n
s
form
er i
s
that a grou
p of coole
r
termi
nal
s
has bu
rn
ed, the hand of th
e on-lo
ad tap
-
ch
ang
er (OL
T
C) i
s
norm
a
l
but the electric op
eratio
n is
bad an
d the
oil gaug
e le
vel is lower t
han no
rmal.
The in
spe
c
ti
on of body o
f
transfo
rme
r
is
norm
a
l.
Table 6. Evaluation of thre
e risk facto
r
s,
priority ran
k
i
ng by traditio
nal RPN
Failure
parts
Severit
y
Occurrence
Detection
RPN Priorit
y
ra
nking
Active part
4
4
1
16
7
Winding
6
4
8
192
3
Core
8
6
8
384
1
OLT
C
6
8
6
288
2
Non-electrical pr
otection
4
4
4
64
5
Cooler s
y
stem
2
6
2
24
6
Bushing
4
6
6
144
4
Tank
2
4
2
16
7
Table 6 sho
w
s the failure ri
sk fa
ctors S, O,
and D for
a powe
r
tran
sformer a
nd h
o
w they
woul
d be
ran
k
ed
usi
ng th
e tradition
al
RPN valu
es.
Clea
rly, the three
risk fa
ctors of p
o
ten
t
ial
failure for the
Active part and the Tank
are differe
nt. Ho
wever, u
s
i
ng tradition
al
RPN cal
c
ulu
s
,
these two pa
rts wo
uld be
assi
gne
d the same
RPN and have the same m
a
in
tenan
ce pri
o
rity
ran
k
ing.
This
pap
er
evaluate
s
th
e intert
wine
d
impor
t
ance
of the
risk factors
of
severity,
occurre
n
ce a
nd dete
c
tion i
n
a po
wer t
r
a
n
sformer,
q
u
antifies the
s
e
qualitat
ive
results plus multi-
expert opini
o
n
s, and
cre
a
tes a jud
g
men
t
matrix as follows:
1
3
4
1
/
3
1
3
1/
4
1/
3
1
D
(11)
(,
,
)
ii
i
i
M
Ex
En
H
e
W
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Failure Mode
and Effect Analysis of Power Tran
sform
e
r Based on
Clou
d… (Jian
peng Bian
)
781
The p
a
ir-wi
s
e
com
p
a
r
ison
clou
d mat
r
ix can
be
co
nst
r
ucted
a
c
cordi
ng to Eq
uatio
n (2
)-(3
)
to determin
e
the cloud m
odel of weig
ht of t
he three variabl
es.
The weig
ht variable
can
be
obtaine
d by i
nputting the
i
ndex value.
F
i
nally, the
cl
o
ud
weight
of
each of the
three
varia
b
le
s i
s
cal
c
ulate
d
utilizing th
e algo
rithm of the
cloud mo
del of
weig
ht. Figure 1 shows th
e clo
ud mo
de
l
of weight of the three
risk factor
s a
c
cord
ing to Equatio
n (4)-(9).
Figure 1. Clo
ud model of
weig
ht of three risk facto
r
s
Table 7. Clo
ud wei
ght of three
risk fact
or
s, p
r
iority ra
nkin
g and ma
intenan
ce
strategy
Failure
parts
Severit
y
Occurrence
Detection
RPN
Priorit
y
ranking
Maintenance stra
teg
y
Active part
4 4
1
1.8310
8
Correc
t
i
v
e
maintenance
0.1366 0.14
04
0.7230
Winding
6 4
8
6.7588
3
Maintain as quickly
as
possible
0.1994 0.21
06
0.5900
Core
8 6
8
7.7184
1
Maintain as quickly
as
possible
0.4353 0.14
04
0.4242
OLT
C
6 8
6
7.2076
2
Maintain as quickly
as
possible
0.1943 0.60
38
0.2019
Non-electrical
protection
4 4
4
4.0004
5
Time-based
maintenance
0.3291 0.33
81
0.3329
Cooler s
y
stem
2 6
2
2.5020
6
Extended
maintenance
0.4443 0.12
55
0.4302
Bushing
4 6
6
5.3258
4
Time-based
maintenance
0.3374 0.32
14
0.3413
Tank
2 4
2
2.2682
7
Extended
maintenance
0.4399 0.13
41
0.4260
Table 7
sho
w
s that the risk value of fail
ure
a
c
ross all
three ri
sk fa
ctors is ve
ry high for
the Core,
O
L
TC
and
Wi
nding
failure
part
s
of
th
e
tran
sform
e
r;
therefo
r
e, th
ese
compo
n
ents
sho
u
ld be m
a
intaine
d
as quickly as
possibl
e a
ccordin
g to the
threshold of
RPN. Two
risk
factors—
occu
rre
nce an
d d
e
tection
—
are
com
paratively high fo
r the
Non
-
ele
c
tri
c
al
prote
c
tion
an
d
Bushin
g p
a
rt
s, but the
se
verity factor i
s
comp
arativ
ely low. T
h
u
s
, potential fail
ure
mainten
a
n
ce
sho
u
ld be time-ba
s
e
d
acco
rding to the thre
shol
d of RPN. One risk factor—o
ccu
rre
nce—i
s
hig
h
for the
Cool
e
r
sy
stem a
n
d
Tan
k
pa
rts,
but the
severi
ty and dete
c
ti
on facto
r
s a
r
e low; th
erefo
r
e,
failure m
a
int
enan
ce fo
r t
hese two
pa
rts should
o
c
cur on
an
extended
maint
enan
ce
sche
dule
according to the thre
shol
d RPN.
Although the
severity an
d occurre
n
ce factors
for the
Active part are com
paratively high,
the detectio
n
factor i
s
very low. Mainte
nan
ce fo
r thi
s
pa
rt, whi
c
h
is behi
nd the
tank, may o
c
cur
on a co
rrectiv
e
sched
ule.
4. Conclusio
n
This p
ape
r illustrate
s the
use of the
cl
oud mo
del of
weight to de
termine the
mutually
intertwin
ed i
m
porta
nce of FMEA failure risk fact
o
r
s occurren
ce
(O), seve
rity (S) and dete
c
tion
(D)
to
analy
z
e a power transfo
rme
r
and optimize
a fin
e
ly tun
ed
risk-adju
s
ted mai
n
tena
nce
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 3, September 20
15 : 776 – 782
782
sched
ule. Th
e pape
r com
pare
s
FMEA
based on
clo
ud wei
ghts to
the curre
n
tly more
comm
only
use
d
tradition
al FMEA, whi
c
h
sco
r
e
s
a
n
d
p
r
ioriti
ze
s
risk through
a
sim
p
le
cal
c
u
l
ation of
RP
N.
The cl
oud m
odel of weig
ht is more practical and
fl
exible than traditional
RP
N value
s
, as it is
cap
able
of ta
king i
n
to
con
s
ide
r
ation
the
relative
im
po
rtance am
on
g the
risk fa
ctors O, S a
nd
D,
as well as all
o
win
g
for un
certai
nties th
at c
an o
c
cur
durin
g tran
sf
orme
r pe
rformance test
s and
evaluation
s
. Based
on the
pra
c
tical exa
m
ple offe
re
d i
n
the pap
er, the clo
ud
wei
ght based mo
del
sho
w
s its p
o
tential adva
n
tage in det
ecting hi
gh
risks of po
wer failures i
n
transfo
rme
r
s
system
aticall
y
and effectively.
Also, very importantly, we
sho
w
that evaluat
ing ri
sk factors u
s
ing
clou
d weig
ht analysi
s
may help p
a
rse o
u
t the a
c
tual differe
nce
s
in ri
sk
that
may lurk be
hi
nd the a
ppa
rently equal
RPN
that traditio
n
a
l
FMEA a
naly
s
is ge
ne
rally
prod
uces
.
It is evid
ent th
at the
pro
p
o
s
e
d
mo
del
ca
n
not
only re
du
ce
manpo
we
r i
n
vestment in
p
o
we
r tran
sf
ormer mainten
a
nce, but
al
so
mitigate
the ri
sks
and expe
nse
s
asso
ciate
d
with po
wer transfo
rme
r
failure
s.
Ackn
o
w
l
e
dg
ements
This
wo
rk
wa
s supp
orted
by the Natio
n
a
l Na
tu
ral S
c
ience Fou
n
d
a
tion of Chin
a (G
rant
No. 5
130
711
2)
and
Natio
nal
Natural S
c
ien
c
e
Fou
n
d
a
tion of
Chi
n
a (Gra
nt No. 512
741
44) a
n
d
the Natu
ral
Scien
c
e F
o
u
ndation
of Hebei Pr
ovince
(G
rant No
.
F2012
210
0
31) and Chi
n
a
Postdo
ctoral Scien
c
e
F
o
u
ndation (G
ra
nt
No.
2
013T
6019
7).And
Colle
ge
s an
d
Unive
r
sitie
s
youth
talent prog
ra
m proje
c
ts of
Heb
e
i Provin
ce (Gra
nt No.
B
J201
405
4).
Referen
ces
[1]
Hose
yn
ab
ad
i
HA, Oraee
H,
T
a
vner PJ. F
a
il
ure M
odes
a
n
d
Effects Anal
ys
is (F
MEA) for
w
i
nd
turb
ines
.
Electrical P
o
w
e
r and En
ergy
Systems
. 20
10
; 32(7): 817-8
2
4
.
[2]
X
i
ao YC, Nan
GQ,
Zhang Qing, Han
Xiao.
T
r
ansformer
Fault Diagnosis
Based on
Hier
archic
al Fuzz
y
Supp
ort Vector
Machin
es.
T
E
LKOMNIKA Indon
esia
n Jo
ur
nal
of El
ectrica
l
Engi
ne
erin
g
. 201
3; 11(1
0
):
584
2-58
50.
[3]
Gao LA, Zhou
YG, Huo LM. Reli
ab
ilit
y asse
ssment of
distributi
on s
y
stem
s
w
i
th
distrib
u
ted
ge
ner
atio
n
base
d
on Ba
ye
sian n
e
t
w
orks.
Engi
neer
in
g R
e
view
. 201
4; 34(1): 55-6
2
.
[4]
Hu AH, Hsu C
W
, Kuo
T
C
,
W
u
W
C
. Risk
eval
uatio
n of g
r
een com
pon
e
n
ts to hazard
o
us substanc
e
usin
g F
M
EA and F
A
HP.
Expert Systems w
i
th Appl
icatio
ns
. 200
9; 36(3): 71
42-7
147.
[5]
Gargama H, C
haturve
di SK. Critical
it
y
Ass
e
ssm
ent Mode
l
s
for F
a
ilure
Mode Effects and Cr
itical
it
y
Anal
ys
is Usin
g
F
u
zz
y
Lo
gic.
Reli
ab
ility
. 201
1; 60(1): 10
2-1
10.
[6]
Bragli
a
M, F
r
osolin
i M, Montanar
i R. F
u
zz
y criticalit
y asse
ssment
mode
l
for failure mo
d
e
s and
effects
analy
sis.
Intern
ation
a
l Jo
urna
l of Quality & Re
liab
ility Ma
nag
ement
. 200
3; 2
0
(4): 503-
52
4.
[7]
F
r
anzén A, K
a
rlsso
n S. F
a
i
l
ure Mo
des
a
nd Effect
s An
al
ysis of T
r
an
sforme
rs, Ro
yal Institute of
T
e
chnolog
y.
K
T
H School of E
l
ectrical E
ngi
ne
erin
g
. Stockhol
m, S
w
ed
en. 2
0
07.
[8]
Liu H
C
, Liu
L, Lin QL. F
u
zz
y F
a
i
l
ure M
o
d
e
an
d Effects Anal
ys
is Usin
g
F
u
zz
y
Evi
dent
ial R
eas
oni
ng
and Be
lief R
u
le
-Based Meth
od
olo
g
y
.
R
e
li
abi
lit
y
. 2013; 62(
1): 23-2
6
.
[9]
Yang
Z
L
, Bon
s
all S, W
a
ng J
.
F
u
zz
y
R
u
l
e
-B
ased
Ba
yes
i
a
n
Reas
on
ing
Ap
proac
h for Pri
o
ritizatio
n
of
Fa
ilu
re
s in
FMEA.
Relia
bil
i
ty
. 200
8; 57(3): 51
7-52
8.
[10]
Che
n
Y,
Ji
L,
Liu Q. A
ppl
icatio
n R
e
sear
c
h
of
F
u
zz
y T
heor
y i
n
PE
T
eachin
g
Eval
uati
on.
TE
L
K
O
M
N
I
K
A
Indon
esi
an Jou
r
nal of Electric
al Eng
i
ne
eri
n
g
.
2013; 1
1
(9): 5
133-
514
0.
[11]
Lee YS, Kim D
J
, Kim JO, Kim H. Ne
w
F
M
ECA Me
thod
ol
og
y Usi
ng Stru
ctural Importan
c
e and F
u
zz
y
T
heor
y
.
Power System
s
. 20
11
; 26(4): 236
4-2
370.
[12]
Liu HC, Li
u L, Liu
N, Mao L
X
. Risk Eval
ua
tion in F
a
il
ure
Mode a
nd Effects Anal
y
s
is
w
i
th Exten
d
e
d
VIKOR Method
under F
u
zz
y E
n
viro
nment.
Power System
s
. 201
2; 39(1
7
): 1292
6-12
93
4.
[13]
Jiao P, Z
ha Y
B
.
Credib
ility
Evalu
a
tion M
e
thod b
a
se
d on
Clou
d
Mod
e
l
. Internati
o
n
a
l C
onfere
n
ce o
n
Electrical and Contro
l
Engi
ne
erin
g. 201
0; 1: 648-
651.
Evaluation Warning : The document was created with Spire.PDF for Python.