TELKOM
NIKA
, Vol.13, No
.2, June 20
15
, pp. 722 ~ 7
2
9
ISSN: 1693-6
930,
accredited
A
by DIKTI, De
cree No: 58/DIK
T
I/Kep/2013
DOI
:
10.12928/TELKOMNIKA.v13i2.1455
722
Re
cei
v
ed
Jan
uary 21, 201
5
;
Revi
sed Ap
ril 8, 2015; Accepte
d
April 2
5
, 2015
Optimal Two Dimensio
nal Preventive Maintenance
Policy Based on Asymmetric Copula Function
Xin
y
ue Li*, Yunxian Jia, Z
h
en Li
Dep
a
rtment of Mana
geme
n
t Engi
neer
in
g,
Mechan
ical En
gi
n
eeri
ng Co
lle
ge
Shiji
azh
u
a
ng, Heb
e
i, Chi
na, Ph.:+
8613
930
1
192
96
*Corres
p
o
ndi
n
g
author, e-ma
i
l
: oecl
x
y@
gmai
l.com
A
b
st
r
a
ct
For some ki
nd
s of prod
ucts, the cons
u
m
ers
have
strict re
quir
e
ments to
the rel
i
ab
ility o
f
thes
e
prod
ucts in t
h
e
base
d
w
a
rra
nty peri
od. T
h
en
the
ma
nuf
actu
rer is i
n
cli
n
e
d
t
o
prov
id
e the
tw
o-dimens
ion
a
l
preve
n
tive
ma
i
n
tena
nce
pol
ic
y to take the
usag
e de
gr
e
e
of the pro
duct
into acco
unt.
As a result, tw
o
-
di
me
nsio
nal p
r
eventiv
e mai
n
tena
nc
e
pol
ic
y in the w
a
rr
anty per
iod
h
a
s rece
ntly o
b
tain
ed i
n
cre
a
s
ing
attention fro
m
ma
nufactur
e
rs
and co
nsu
m
e
r
s. In th
is pap
er, w
e
focuse
d on th
e opti
m
i
z
at
io
n of ba
sed
w
a
rranty cost
and pr
opos
ed
a new
expecte
d base
d
w
a
rrant
y cost mod
e
l
consid
erin
g the tw
o-dime
nsi
o
na
l
imperfect
prev
entive
mai
n
ten
ance
p
o
licy
from the
p
e
rsp
e
ctive of
the
ma
nufactur
e
. Asymmetric
co
pul
a
function w
a
s a
ppli
ed to
mod
e
lin
g the fa
ilur
e
functi
o
n
of the pr
oduct. A
nd the
opti
m
a
l
tw
o-dime
nsi
o
nal
preve
n
tive
mai
n
tena
nce
peri
od w
a
s o
b
tai
n
ed by
mi
n
i
mi
zing
bas
ed w
a
rranty cost. At last, nu
meric
a
l
exa
m
p
l
es are
give
n
to ill
ustrate
the
pr
opos
ed mode
ls,
of
w
h
ich the
res
u
lts prov
e th
e
mode
l effectiv
e
an
d
valid
ate.
Ke
y
w
ords
: Ba
sed W
a
rranty, T
w
o-Dime
n
sio
nal Prev
entiv
e
Mainte
nanc
e, Asymmetric C
o
pul
a F
unctio
n
1. Introduc
tion
In general, warranty has two sig
n
ifica
n
t roles
a
s
prot
ector to consumers and p
r
omoter t
o
manufa
c
turer whe
n
the
produ
cts fail to
perfo
rm t
hei
r pre
-
spe
c
ified function
s
d
u
ri
ng
the wa
rra
n
ty
perio
d. Acco
rding to Murth
y
and Blisch
ke [1], ther
e are only two kin
d
s of wa
rra
nty in the produ
ct
life
cycle
i
n
cl
uding ba
sed warra
n
ty
and extended wa
rr
anty. Base
d
warra
n
ty is
sold with
prod
uct.
Its co
st is u
s
ually inclu
ded
in the sale
p
r
ice.
Ho
weve
r, to have extende
d wa
rra
n
ty, the custo
m
er
sho
u
ld pay e
x
tra money to get the service afte
r b
a
s
ed
wa
rra
nty expire
s. No
wad
a
ys, eith
er for
based warra
n
ty or exten
ded warranty
,
one key i
s
s
ue
of intere
st that ari
s
e
s
from
wa
rra
nty
analysi
s
is th
e modelin
g o
f
warranty co
st. By
optimizing wa
rranty co
st, the manufactu
re
r co
ul
d
price the wa
rranty pro
per
and ma
ke th
eir produ
ct
more attra
c
ti
ve to consu
m
ers. One of
th
e
possibl
e
ways to
achieve
the ab
ove g
o
a
l is by
ma
ki
ng o
p
timal d
e
ci
sion
on
th
e mainte
nan
ce
strategi
es in
the warranty
peri
od. In t
he literat
ure, mainten
a
n
c
e is mainly
classified int
o
two
types: corre
c
tive mainte
na
nce
(CM
)
a
n
d
p
r
eventiv
e
maintena
nce
(PM) [2]. If th
e u
s
eful
life o
f
a
prod
uct i
s
rel
a
tively sho
r
t, like la
mp
bul
b, then it
s wa
rra
nty is
also
relatively
sh
ort an
d
warra
n
ty
servi
c
ing should involve
only CM
actions. If
a
pro
d
u
ct h
a
s a l
o
ng u
s
eful
life or
nee
ds hi
gh
availability, like milita
r
y e
quipme
n
t, then its
wa
rr
an
ty is rel
a
tively long a
nd
manufa
c
turer ca
n
redu
ce
wa
rranty servi
c
in
g co
sts a
n
d
improve
its perfo
rman
ce by perfo
rming effectiv
e PM
action
s. Mai
n
tenan
ce i
s
th
erefo
r
e
signifi
cant in
the
warranty co
nte
x
t since
it ha
s a m
a
jo
r imp
a
ct
on expe
cted
warra
n
ty serv
icing
co
st.
For s
o
me k
i
nds
of
c
o
mplex
produc
t
s
(s
uc
h
as
aircraft, government acquis
i
tion), the
con
s
um
ers h
a
ve
st
rict req
u
irem
ents
to the
reliab
ility
of these p
r
o
d
ucts. S
o
the
manufa
c
turer is
inclin
ed to
provide the
two
-
dime
nsi
onal
preventiv
e
m
a
intena
nce p
o
licy to ta
ke
the u
s
a
ge
deg
re
e
of the p
r
o
d
u
c
t into a
c
coun
t, which h
a
s
two limit
s
of coverage
in
t
e
rm
s
of age
and usage. F
o
r
example, a
sold
car may
be
covered
a
3 yea
r
a
nd 3
0
,000 mil
e
warranty, and t
he
car shoul
d
be
che
c
k u
p
in
4
S
(Sale, Sp
are pa
rt, Servi
c
e an
d
Su
rvey)
store
at p
e
r
year a
nd
15,
000 mil
e
. F
r
o
m
the pe
rspe
ctive of the
co
nsum
er
s, two-dim
e
n
s
iona
l preve
n
tive maintena
nce
policy mi
ght
be
better than o
ne-di
men
s
ion
a
l warra
n
ty for its on
e
mo
re bou
nda
ry that coul
d ma
ke the p
r
od
u
c
ts
be maintai
n
e
d
in time. Ho
wever, it coul
d al
so
pla
c
e
heavier
burd
en on the m
anufa
c
ture
rs. So
the
manufa
c
t
u
re
rs sh
ould price
the warranty
prop
erly
to make th
eir
prod
uct
s
mo
re attractive a
nd
guarantee
th
eir p
r
ofit. As a result, how to
minim
i
zing th
e two-dim
e
n
s
iona
l wa
rra
nty cost
con
s
id
erin
g two-dime
nsio
nal PM be
co
mes a
pra
c
ti
cal requi
rem
ent. To our
kno
w
le
dge, fe
w
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Optim
a
l Two Dim
ensio
nal
Preventive Maintena
nce
Policy
Bas
e
d on As
y
mmetric ...
. (Xiny
ue Li)
723
analytical
or
nume
r
ical m
e
thod
s h
a
ve
bee
n repo
rt
ed in
the lite
r
ature
so fa
r,
whi
c
h
offers a
deci
s
io
n fra
m
ewo
r
k to o
p
timize th
e t
w
o-dime
nsio
nal p
r
eventiv
e mainte
nan
ce
peri
od in
the
warra
n
ty peri
od. In this pe
rsp
e
ctive, we
develop
in th
is pap
er a
ne
w mathe
m
atical mo
del to find
the optimi
z
e
d
preventive
mainten
a
n
c
e poli
c
y fo
r ba
sed
wa
rranty from th
e pe
rspe
ctive of
manufa
c
turer. The rest of this pape
r is
orga
nized a
s
follows: Sect
ion 2 introdu
ce
s the relati
ve
work existin
g
in the literature with p
r
o
duct
warrant
y policie
s. Section 3 outli
nes the mo
d
e
l
assumptio
n
s
and notatio
n. Section 4 i
s
d
edicated
to d
e
velopme
n
t of the mathematical mod
e
l. A
nume
r
ical ex
ample
illust
rates
ou
r a
p
p
roa
c
h
to
prove the
mo
del valid
ate
and
effective
in
Section 5.
2. Literature
Rev
i
e
w
Different warranty policie
s has different warr
a
n
t
y
cost
.
The cha
r
a
c
t
e
rist
i
cs of
w
a
rra
nt
y
polici
e
s are
essential
fact
ors in
dete
r
mining
wa
rr
a
n
ty co
st in
cl
uding
warran
ty servi
c
e
ob
ject,
maintena
nce
task,
warra
n
ty ren
e
wal
mech
ani
sm
and wa
rranty
dime
nsio
n etc.
Blisch
ke
an
d
Murthy [3] formulate
d
a classic taxono
my for
warra
n
ty policies.
But when th
e prod
uct
s
a
r
e
becoming
mo
re
com
p
lex, t
he
wa
rra
nty p
o
licie
s
hav
e
some
ch
ange
s. So ba
se
d o
n
the
sy
stem
of
the wa
rra
nty polici
e
s [4]-[6
], we further
develop t
he t
a
xonomy for
warra
n
ty policie
s to revie
w
the
curre
n
t literature
s
, sho
w
n
as Figu
re 1.
Figure 1. The
classification
of warranty policie
s
At first, wa
rranty polici
e
s
coul
d be
divi
ded into
two
types in
clu
d
i
ng si
ngle
-
co
mpone
nt
warra
n
ty and
multi-comp
o
nent warrant
y. The forme
r
mainly fo
cu
se
s on th
e
warranty poli
c
y for
singl
e comp
o
nent o
r
syste
m
[7]-[8]. On
the co
nt
ra
ry, multi-compo
n
ent wa
rranty
often consi
d
e
r
s
a fleet
of produ
cts. T
hen
the
wa
rra
nt
y coul
d
b
e
divided
i
n
to two sub
-
g
r
ou
ps as
rene
wing
warra
n
ty and
non-re
ne
wing
wa
rra
nty. Re
newi
ng
warr
a
n
ty policy me
ans th
at the
warra
n
ty peri
o
d
will restart at the failure time if the failure o
ccurs in the warranty peri
od. Renewing warranty cost
is rel
a
tively difficult to be cal
c
ulate
d
. Bai pre
s
e
n
ts cost mod
e
ls f
o
r re
pai
rable
multi-compo
n
ent
system
s [9]
and a
nd Pa
rk propo
se
d the re
ne
wing
warra
n
ty co
st model [1
0
]. Non-ren
e
wing
warra
n
ty policy will end at
the warranty
period
W no
matter how
many failure
s happe
n in the
warra
n
ty peri
od [11]-[1
2
]. Simple warra
n
ty polic
y ha
s only
one
warranty poli
c
y will be
ap
plie
d in
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 722 – 72
9
724
the warranty
period [13], and differ
ent
warranty poli
c
i
e
s
will be app
lied in the period of
com
b
ined
warra
n
ty policy [14]. Accordin
g to the
warrant
y bo
unda
rie
s
, the
warra
n
ty policie
s co
uld
be
further
divide
d into one
-di
m
ensi
onal
warranty,
two-dimen
s
ion
a
l
warra
n
ty and
n-dime
nsi
o
n
a
l
war
r
a
n
ty
(n
≥
3). On
e-dime
nsio
nal
wa
rra
nty is limited
by
time [15],
and t
w
o-dime
nsio
nal
wa
rra
nty
is u
s
ually limi
t
ed by age a
n
d
usage
(such as mil
e
s
an
d revolutio
n
s) [16]-[18]. For some
co
mpl
e
x
prod
uct
s
, whi
c
h requi
re hi
gh relia
bility, multi usa
ge li
mits will be a
pplied to rest
rain the
warranty
perio
d. The
n
the wa
rranty
coul
d be
n-di
mensi
onal
warranty. There are fe
w literatures
abo
ut n-
dimen
s
ion
a
l
warra
n
ty. Ho
wever, al
ong
with the ind
u
strial
pro
d
u
c
t be
come m
o
re
compl
e
x, the
intere
sting
of
the n
-
dime
nsi
onal
wa
rranty wo
uld
in
cre
a
se.
Obviou
sly wa
rranty p
o
licy is the
resu
lt
of the game between ma
nufacturer
and consumer. So
the co
st allocation
mechanism will
greatly
influen
ce the
warra
n
ty deci
s
ion
and
effe
ct. And
the
warranty poli
c
y coul
d b
e
furt
her
divided i
n
to
free-rep
a
ir
warranty, pro-rate wa
rra
nty and rebat
e
warranty etc b
y
cost allo
cat
i
on me
chani
sm.
The manufacturer will take all ex
pense i
n
the free-repai
r
warranty peri
od. On
the contrary, the
warranty cost
will
be share by
manufacturers and
consum
ers i
n
the period
of pro-rata warranty
or re
bate warranty [19].
Obliviou
s
ly, two-dime
nsio
nal
warranty i
s
o
ne
of
the
main
conte
n
ts of th
e
warra
n
ty policy
For two
-
dime
nsio
nal wa
rra
n
ty, there are
three
metho
d
s that has b
een develo
p
e
d
for analyzin
g
two-di
men
s
io
nal warranty
[20], ie univa
riate m
e
t
hod
[21]-[23], biv
a
riate
metho
d
[24] an
d ti
me
scale metho
d
[25]
。
Bivairate metho
d
is sim
p
ler a
n
d
more
strai
ghtforward. And one
wa
y to
formulate
the
bi-va
r
iant
s di
stributio
n i
s
makin
g
fu
ll
u
s
e
of Copul
a
function.
Co
p
u
las are a
too
l
for
con
s
tru
c
ting
multivariate
dist
ribution
s
and
de
scri
bing th
e d
e
pend
en
ce
b
e
twee
n
rand
om
variable
s
F
u
rthermo
re
symmetric co
p
u
la fun
c
tion
were al
ready
applie
d to
model the
produ
ct
reliability. In
this pap
er,
we
p
r
op
ose the
cost
model
of th
e two
-
dim
e
n
s
ion
a
l p
r
eve
n
tive
maintenance in the warranty period by
utilizing
asymmetric copula
func
tion. And the opti
m
al
two-di
men
s
io
nal preve
n
tive maintena
n
c
e pe
riod i
s
obtaine
d by minimizi
ng the warranty cost.
The advanta
g
e
s of the mod
e
l prop
osed a
r
e as follo
ws.
-
It is able to
handl
e the t
a
il dep
end
en
ce of
age
a
nd u
s
ag
e to
cal
c
ulate
two-dim
e
n
s
iona
l
warra
n
ty cost which
co
uld be mo
re a
c
curate. Th
e
asymmetri
c
cop
u
la fun
c
tion is ap
plie
d
rathe
r
than a
s
suming the
age an
d usag
e have linea
r relation
shi
p
.
-
The t
w
o-dime
nsio
nal
preve
n
tive mainte
n
ance
p
e
rio
d
could be obtai
ned by
minim
i
zing
ba
se
d
warra
n
ty co
st. It is more realisti
c that t
he PM
a
c
tio
n
is
also two
-
dime
nsi
onal
to make th
e
prod
uct mo
re
reliable.
3. Model con
s
ideratio
n
This se
ction provide
s
m
o
del
con
s
ide
r
a
t
ion
and
som
e
prelimina
r
y re
sults. T
h
rough
out
this pa
pe
r, we assu
me im
perfe
ct preve
n
tive mainten
ance an
d mi
nimal
corre
c
ti
ve mainten
a
n
c
e.
Comp
ared to the mean time betwe
en failure
s, the
maintena
nce time is negligi
b
le. A failure can
be
d
e
tecte
d
immediately whi
c
h re
sults
in
a
n
imm
e
diate
claim,
then the
ma
nufactu
re
r
wi
ll
respon
d all t
he
claim
s
. Assuming
a fa
ilure
co
uld b
e
dete
c
ted i
mmediately
whi
c
h
re
sults in
a
n
immediate
cl
aim and the
manufa
c
turer will respon
d all the claim
s
. Notations a
s
follow:
(T
0
, U
0
)
: The two-di
m
ensi
onal p
r
ev
entive mainte
nan
ce (PM
)
p
e
riod im
plem
ented
(W
B
, U
B
)
: The based
warra
n
ty time and usage li
mit and
0
0
0
B
B
U
U
r
WT
(0
1
)
: PM level.
When
1
, PM is pe
rfect ma
intenan
ce. When
0
, PM is
minimal re
pai
r.
C
p
: The co
st of PM
C
m
: The co
st of CM
Assu
ming th
e
usa
ge of th
e
prod
uct i
s
rel
a
tively steady
for individ
ual
cu
stome
r
. It doe
sn’t
mean th
e u
s
age fo
r in
dividual
cu
stome
r
do
esn’t
cha
nge but cha
n
ge
little.
The
based wa
rra
n
ty
coverage a
r
e
a
coul
d be di
vided into two
parts: D
1
a
nd D
2
sho
w
n a
s
Figure 2. Fo
r D
1
, warranty is
likely terminat
ed at usag
e limit. On the contra
ry,
warranty is likely termin
ated at time limit for
D
2
.
In this stu
d
y, the wa
rrant
y cost i
s
co
mposed
of p
r
eventive mai
n
tenan
ce
co
st and corre
c
tive
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Optim
a
l Two Dim
ensio
nal
Preventive Maintena
nce
Policy
Bas
e
d on As
y
mmetric ...
. (Xiny
ue Li)
725
maintena
nce
cost. The cumulative failure di
stri
b
u
tion functio
n
, failure de
nsit
y functions a
n
d
failure rate of the prod
uct a
r
e [20]:
12
01
2
1
1
2
2
1
2
2
21
3
1
2
3
(,
)
(
(
)
,
(
)
,
)
(
(
),
(
)
,
)
(
1
,
(
),
)
(
1
(
),
(
)
,
)
((
)
,
1
,
)
(
(
)
,
1
(
)
,
)
,
(,
)
(,
)
(,
)
,
(,
)
1
(
,
)
Ft
u
C
F
t
F
u
p
C
F
t
Fu
p
C
Fu
C
F
t
F
u
pC
F
t
C
F
t
F
u
Ft
u
Ft
u
ft
u
h
t
u
F
t
u
tu
tu
(1)
whe
r
e
12
((
)
,
(
)
,
)
CF
t
F
u
is the
asymmet
r
ic cop
u
la fun
c
tion a
nd
2
0
1
i
i
p
.
12
1
((
)
,
(
)
,)
CF
t
F
u
is
symmetri
c
co
pula
fun
c
tion
and com
p
o
s
e
d
by
sy
mm
etric copul
a. Th
e co
mmon
si
mple
symmet
r
ic
cop
u
la functi
ons in
clud
e Gau
ssi
an co
pula, t-co
pula
,
Gumbel co
p
u
la, Clayton cop
u
la and F
r
an
k
cop
u
la etc. G
u
mbel copul
a
is:
1/
'
12
1
2
((
)
,
(
)
,
)
e
x
p
(
l
n
((
)
)
(
l
n
(
(
)
)
CF
t
F
u
F
t
F
u
(2)
Asymmetri
c
copula fun
c
tio
n
coul
d ha
ndl
e tail depe
nd
ence betwee
n
time and u
s
age in a
given di
rection, whi
c
h
coul
d be
applied i
n
modeling
reliability
data.
Based on the collected dat
a
,
we could e
a
si
ly estimate the para
m
eters of
12
()
,
(
)
F
tF
u
whi
c
h a
r
e
margin
al dist
ribution
s
.
Figure 2. Two
dimensi
onal
warra
n
ty coverag
e
4. Analy
t
ical
w
a
r
r
an
t
y
co
st analy
s
is
For th
e two
-
dime
nsi
onal
impe
rfect
preventive
p
o
licy, the fa
ilure
rate
ri
ght after
perfo
rming
th
e jth PM a
c
tio
n
is
given by
Eq. (4).
Obvi
ously, the PM
action
effect
is the
red
u
cti
on
of failure intens
ity [26].
10
B
nU
U
,
20
B
nW
T
(
*
:
Integer part of a real number)
.
00
0
0
0
00
0
0
0
PM
a
t
(
,
)
(
1
)
(
,
)
PM
a
t
(
t
,
)
(
1
)
(
/
,
)
jT
h
j
T
u
h
j
T
j
T
r
j
U
h
jU
h
j
U
r
jU
j
=
1
,2,…
(3)
For t
w
o-dime
nsio
nal
warranty, there a
r
e two
po
ssi
b
le situ
ation
s
that ba
sic
warra
n
ty
expire
s. One
is the usage
of
the produ
ct excee
d
s th
e mileage lim
itation. And the othe
r is th
e
age
of the
produ
ct exceed
s the
time
lim
itation.
So th
e warranty
co
st in
clud
es two pa
rts:
D
1
an
d
D
2
. For the warranty cove
rage of D
1
, let N
t
be the nu
mber of PM o
b
se
rved up to
time t. Then the
failure rate be
tween two PM action is:
t
N
0
,
(,
)
ht
u
;
t
N
1
,
00
0
(,
)
(
,
)
ht
u
h
U
r
U
;
…
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 722 – 72
9
726
t
N
j
,
1
0
0
0
0
()
(,
)
(
1
)
,
(
)
j
i
i
ji
U
ht
u
h
j
i
U
r
The co
st of the warra
n
ty includi
ng two
-
di
mens
i
onal im
perfe
ct preve
n
tive mainten
ance is:
00
1
1
0
0
1
10
(1
)
1
1
0
1
0
00
0
0
1
10
10
0
0
0
()
(,
)
(
1
)
,
(
)
()
(,
)
(
1
)
,
(
)
B
jU
u
r
n
j
i
Dp
m
ji
jU
ur
U
n
i
m
i
nU
ji
U
C
C
n
C
h
t
u
h
j
i
U
d
tdtu
r
ni
U
Ch
t
u
h
n
i
U
d
t
d
t
u
r
(4)
At the same
prin
ciple, th
e
co
st of th
e
warra
n
ty
incl
uding
two
-
di
mensi
onal
im
perfe
ct p
r
eve
n
tive
maintena
nce for the wa
rra
nty coverag
e
of D
2
is
:
00
2
2
0
0
2
20
(1
)
1
1
2
00
0
00
0
1
20
20
0
0
0
(,
)
(
1
)
(
)
,
(
)
(,
)
(
1
)
(
)
,
(
)
B
jT
r
t
n
j
i
Dp
m
ji
jT
rt
W
n
i
m
i
nT
CC
n
C
h
t
u
h
j
i
T
j
i
T
r
d
u
d
t
C
h
t
u
h
n
i
T
n
i
T
r
du
dt
(5)
So the total expected b
a
se
d warra
n
ty co
st is
1
2
1
2
12
12
(,
)
(
,
)
()
(,
)
(
,
)
(,
)
(
,
)
DD
BD
D
DD
DD
f
t
u
d
t
d
u
f
t
u
dt
du
EC
C
C
f
t
u
dt
du
f
t
u
d
t
d
u
f
t
u
dt
du
f
t
u
d
t
d
u
(6)
5. Numerical
example
We
u
s
e
th
e asymmet
r
ic copula pro
p
o
s
ed
by
Wu [2
0]. The sym
m
etric
co
pula
function
whi
c
h is a
ppli
ed to con
s
tru
c
tion the a
s
ymmetric
cop
u
l
a is
12
1
2
0
1
2
(
(
)
,
()
)
(
)
(
)
1
(
1
(
)
,
1
()
)
CF
t
F
u
F
t
F
u
C
F
t
F
u
.
whe
r
e
1/
0
(
,
)
e
x
p
ln
(
)
ln
(
)
Cx
y
x
y
is the
Gumbel
cop
u
la. Suppo
se
p
C
150,
m
C
820 an
d
0.36. The
ba
sed
wa
rranty and exten
ded
wa
rra
nty cov
e
rag
e
a
r
e (5,
1000
00) an
d
(1
0, 20
000
0
)
. The
life
of the p
r
o
d
u
c
t
is(3
0, 60
000
0).
0
0.9
p
,
1
0.1
p
,
1
3.
76
and
2
0.
48
,
1
0.7
8
,
2
0.77
,
1
2.6
,
2
48
40
0
. The margi
nal cum
u
lati
ve failure
distrib
u
tion
s are:
1
1
1
()
1
e
x
p
t
Ft
,
2
2
2
()
1
e
x
p
u
Fu
In the ba
sed
warra
n
ty, the optimal p
r
eve
n
ti
ve mainten
ance pe
riod
could b
e
obtai
ned by
min
i
mize
th
e
b
a
s
e
d
wa
rra
nty
shown
as
Figure
3. Th
e optimal warranty cost is
513.1398 wh
en
the preventive maintenance period is (1.2, 24000).
Pa
rtial results of the F
i
gure 3 are shown
as
Table 1. We
could find the warranty cost will decreas
es at first then increases
with the increasing
of
T
0
or
U
0
.
Then we anal
yze the sensi
t
ive of th
e
ba
sed
warranty cost mod
e
l b
y
changing
shown
as Figure 4. We could notice some interesting pheno
mena. Warra
nty cost of different PM inte
rva
l
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Optim
a
l Two Dim
ensio
nal
Preventive Maintena
nce
Policy
Bas
e
d on As
y
mmetric ...
. (Xiny
ue Li)
727
change mo
re
by increasin
g
1
and
2
. And the optimal warranty cost is increa
sing
when
1
is
increa
sing. O
n
the contra
ry, it is decrea
s
ing wh
en
2
is increa
sing.
W
e
come t
o
t
h
e con
c
lusion
that
1
and
2
have opposite influence on the based wa
rran
t
y cost.
0
1
2
3
4
5
0
5
10
x 1
0
4
500
600
700
800
900
1000
1100
Figure 3. Warranty cost curve of different
two-dimensio
nal preventive maintenance period
0
1
2
3
4
5
0
2
4
6
8
10
x 1
0
4
40
0
60
0
80
0
10
00
12
00
12
0.
58
,
0
.77
5
0
5
.
06
59
(
2
.9
,
5
800
0)
B
Mi
n
C
a
t
0
1
2
3
4
5
0
5
10
x 1
0
4
50
0
60
0
70
0
80
0
90
0
10
00
11
00
12
0.
98
,
0
.7
7
512.7
420
(
1.3
,
2600
0)
B
MinC
at
0
1
2
3
4
5
0
5
10
x 1
0
4
400
600
800
1000
1200
1400
12
0.
78
,
0
.57
55
3.5
0
6
6
(
2
.9
,
5
80
00
)
B
M
i
nC
at
0
1
2
3
4
5
0
5
10
x 1
0
4
30
0
40
0
50
0
60
0
70
0
80
0
90
0
12
0.
78
,
0
.9
7
39
4.50
59
(
1
.2
,
2400
0)
B
Mi
nC
at
Figure 4. Sen
s
itive analyze
of based warr
anty cost model considerin
g different
Then we
anal
yze the sen
s
i
t
ive of the
based
warranty cost model
b
y
changing
shown
as Figure 5.
And the optimal wa
rranty cost is decreasing when
1
and
2
is increasing. We
come to the c
onclusion that
1
and
2
have th
e uniform influence on the based wa
rran
t
y cost.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 16
93-6
930
TELKOM
NIKA
Vol. 13, No. 2, June 20
15 : 722 – 72
9
728
Table 1. Warranty cost of d
i
fferent two-
di
mensi
onal p
r
eventive maintenan
ce p
e
ri
od
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
1000
0
1000.
7
725.8
701.2
708.8
715.2
720.4
749.8
772.5
790.9
2000
0
835.2
560.3
535.6
543.3
549.6
554.8
584.3
607.0
625.4
3000
0
820.4
545.5
520.8
528.5
534.8
540.0
569.4
592.2
610.5
4000
0
825.0
550.1
525.5
533.2
539.5
544.7
574.1
596.9
615.2
5000
0
828.8
554.0
529.3
537.0
543.3
548.5
577.9
600.7
619.0
6000
0
832.0
557.1
532.4
540.1
546.4
551.6
581.1
603.8
622.2
7000
0
849.7
574.8
550.2
557.9
564.2
569.4
598.8
621.6
639.9
8000
0
863.4
588.6
563.9
571.6
577.9
583.1
612.5
635.3
653.6
9000
0
874.5
599.6
574.9
582.6
589.0
594.2
623.6
646.3
664.7
0
1
2
3
4
5
0
5
10
x 1
0
4
600
800
1000
1200
1400
1600
12
1.
6
,
48
400
78
8.6
546
(
2
.1
,
420
00)
B
MinC
at
0
1
2
3
4
5
0
5
10
x 1
0
4
400
600
800
1000
1200
12
3.6
,
4
840
0
55
6.341
4
(
2
.
8
,
560
00)
B
MinC
at
0
1
2
3
4
5
0
5
10
x 1
0
4
800
1
000
1
200
1
400
1
600
12
2.6
,
2
8400
902.9131
(3
,
60000)
B
Mi
n
C
at
0
1
2
3
4
5
0
5
10
x 1
0
4
400
600
800
1
000
1
200
1
400
12
2
.
6
,
68
40
0
58
4.7
084
(
2
.
9
,
5
8
000
)
B
Mi
n
C
at
Figure 5. Sen
s
itive analyze
of based warr
anty cost model considerin
g different
6. Conclusio
n
s
This a
r
ticle propo
se
s a method to find optim
al two-di
mensi
onal p
r
eventive maintenan
ce
perio
d by mi
nimizin
g
the
warra
n
ty co
st base
d
on
asymmet
r
ic
copula fu
nctio
n
. The meth
od
prop
osed
cou
l
d help the m
anufa
c
ture
r to pri
c
e t
he b
a
s
ed
wa
rra
nty prop
erly. Man
y
extension
s
to
this wo
rk may be con
s
i
dere
d
. The
method of fa
ilure redu
ctio
n model
s ha
ve to be further
investigate
d
. And the pa
ra
meters of co
p
u
la functio
n
have to be theoretically
stu
d
ied. In ord
e
r to
che
c
k model
s validity, some good
ne
ss-of
-
fi
t test
s sho
u
ld be d
e
velope
d. At last, imperf
e
ct
corre
c
tive ma
intenan
ce
sh
ould be in
clu
ded in the
s
e
model
s.
U
0
T
0
Cost
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
1693-6
930
Optim
a
l Two Dim
ensio
nal
Preventive Maintena
nce
Policy
Bas
e
d on As
y
mmetric ...
. (Xiny
ue Li)
729
Referen
ces
[1]
Blischk
e
W
R
, Murth
y
DNP
.
Product
w
a
rr
ant
y
ma
na
ge
ment-I:
A
t
a
xo
nom
y
for
w
a
rr
ant
y
pol
icies.
Europ
e
a
n
Jour
nal of Operati
o
nal R
e
searc
h
. 199
2; 62(2): 12
7-14
8.
[2]
Moham
ed BD,
Sali
h OD,
Abdu
l R.
H
and
book
of Mai
n
tena
nce M
ana
ge
me
nt an
d
Engi
neer
in
g
.
S
p
ring
er
. 200
5.
[3]
Blischke WR,
Murth
y
DNP
.
W
a
rranty Cost
Analys
is
. CRC
Press. 1993.
[4] Blischke
WR
, Murthy
DNP
.
Product W
a
rra
nty Hand
bo
ok
. CRC Press. 199
5.
[5] Murthy
DNP
,
Blischke WR.
W
a
rranty ma
na
ge
me
nt and pr
od
uct ma
nufactur
e
. S
p
ring
er
. 2006.
[6]
Shafie
e M, Chukova S. Main
tenanc
e m
ode
l
s
in
w
a
rra
nt
y
:
A
literature revie
w
.
Eur
ope
an
Journal of
Operatio
nal R
e
search
. 20
13;
229(
3): 561-
57
2.
[7]
Chuk
ova S,
Ar
nol
d R, W
a
ng
DQ. W
a
rrant
y ana
l
y
sis:
An
appr
oach
to m
ode
lin
g im
perf
e
ct rep
a
irs.
Internatio
na
l Journ
a
l of Prod
uction Eco
n
o
m
ics
. 2004; 8
9
(1
): 57-68.
[8]
Hua
ng
YS, Z
huo
YF
. Estimation of future break
do
w
n
s to determine o
p
timal
w
a
rr
ant
y
polic
ies fo
r
prod
uct
s
w
i
t
h
d
e
terior
ation.
R
e
lia
bi
lity Engi
n
eeri
ng & System Safety
. 200
4
;
84(2): 163-1
6
8
.
[9]
Bai J, Pham
H. Cost anal
ysis on rene
w
a
ble fu
ll-s
e
rvice
w
a
rra
nties for multi-compon
ent s
y
stems
.
Europ
e
a
n
Jour
nal of Operati
o
nal R
e
searc
h
. 200
6; 168(
2): 492-5
08.
[10]
Park M, Pha
m
H.
Altered quas
i-ren
e
w
a
l
conc
e
p
t
s
for mode
lin
g ren
e
w
a
b
l
e
w
a
rr
ant
y cost
s
w
i
th
imperfect rep
a
irs.
Mathematic
al an
d Co
mp
ut
er Mode
lli
ng
. 2
010; 52(
9): 143
5-14
50.
[11]
Bai J, Pham
H. Rep
a
ir-l
imit
risk-fr
ee
w
a
rr
ant
y
p
o
lic
ies
w
i
th imperfect rep
a
ir
.
IEEE T
r
ansactions
on
Systems, Man
and Cy
bern
e
tic
s
, Part
A: Systems a
nd H
u
ma
ns
. 2005; 3
5
(6)
:
765-77
2.
[12]
W
u
S, Long
hu
rst P
.
Optimising a
ge-re
plac
ement
a
nd e
x
t
end
ed n
on-r
e
n
e
w
i
n
g
w
a
rr
ant
y p
o
lic
ies i
n
lifec
ycle c
o
stin
g.
Internatio
nal
Journa
l of Productio
n
Econ
o
m
ics
. 20
1
1
; 13
0: 262–
26
7.
[13]
Park M, Pham H. W
a
rranty
c
o
st ana
l
y
s
e
s
using q
u
a
s
i-rene
w
a
l pr
o
c
esses for multicom
pon
ent
s
y
stems.
IEEE
T
r
ansactio
n
s o
n
Syste
m
s, Ma
n an
d
Cy
bern
e
t
ics, Part
A: Sy
stems
and
Hu
ma
ns
. 20
10;
40(6): 13
29-
13
40.
[14]
Jack N, Iskand
ar B P
,
Murthy
D N P
.
A
rep
a
ir–rep
lac
e
strateg
y
bas
ed o
n
usag
e rate for items sol
d
w
i
t
h
a t
w
o
-
d
i
m
ensi
ona
l
w
a
rra
nt
y
.
Rel
i
a
b
il
ity Engi
neer
in
g & System Safety
.
2009; 9
4
(2): 6
1
1-61
7.
[15]
Park M, Pham H.
A
ne
w
w
a
rrant
y p
o
lic
y
w
i
th fai
l
ure
times and
w
a
r
r
ant
y
serv
icin
g
times.
IEEE
T
r
ansactio
n
s o
n
Reli
ab
ility
. 20
12; 61(3): 8
22-
831.
[16]
Che
n
T
,
Popova E. Mainten
ance p
o
lic
ies
w
i
t
h
t
w
o
-
dim
e
nsio
nal
w
a
rran
t
y
.
Reli
abi
lity Engi
neer
in
g &
System
Safety
.
2002; 7
7
(1): 6
1
-69.
[17]
Chuk
ova S, Johnston MR.
T
w
o-
dime
nsi
ona
l w
a
rrant
y
r
e
p
a
ir
strateg
y
base
d
on minimal a
nd compl
e
te
rep
a
irs.
Math
e
m
atic
al a
nd co
mp
uter mod
e
ll
i
n
g
. 200
6; 44(1
1
): 1
133-1
1
4
3
.
[18]
V
a
rnos
afader
a
n
i S, Chukova
S.
A
t
w
o-dim
e
n
s
ion
a
l
w
a
rra
nt
y servicing strateg
y
bas
ed on r
educti
on in
prod
uct failure
intens
it
y
.
Co
mp
uters & Mathe
m
atics w
i
th
Ap
plicati
ons
. 2
0
1
2
; 63(1): 20
1-2
13.
[19]
Chie
n
YH.
T
he ef
fect of a pro
-rat
a
rebat
e
w
a
rr
ant
y
on th
e a
ge re
plac
eme
n
t
polic
y
w
i
t
h
sa
lvag
e val
u
e
consi
derati
on.
IEEE T
r
ansactions
on Reliability
. 2010; 59(2):
383-3
92.
[20]
W
u
S. Constru
c
tion of as
ymm
e
tr
ic copu
las a
nd it
s ap
plic
ati
on in t
w
o
-
dim
e
nsio
nal r
e
li
abi
li
t
y
mod
e
ll
ing.
Europ
e
a
n
Jour
nal of Operati
o
nal R
e
searc
h
. 201
4; 238(
2): 476-4
85.
[21]
Y
un W
Y
, Kang KM. Imperfect rep
a
ir
po
lici
e
s under t
w
o-
di
mensi
ona
l
w
a
r
r
ant
y
.
Proc
eed
ings of th
e
Institution of Mecha
n
ica
l
Eng
i
neers,
Part O:
Journ
a
l of Risk
and Re
lia
bil
i
ty
. 2007; 2
21(4):
239-
247.
[22]
Iskandar BP
, Murth
y
D
N
P
,
Jack N.
A
new
re
p
a
ir
–re
p
la
ce strateg
y
f
o
r items sold
w
i
t
h
a t
w
o-
dime
nsio
nal w
a
rrant
y
.
C
o
mp
u
t
ers & Operatio
ns Rese
arch
. 2
005; 32(
3): 669
-682.
[23]
Hua
ng
YS, Z
huo
YF
. Estimation of future break
do
w
n
s to determine o
p
timal
w
a
rr
ant
y
polic
ies fo
r
prod
uct
s
w
i
t
h
d
e
terior
ation.
R
e
lia
bi
lity Engi
n
eeri
ng & System Safety
. 200
4
;
84(2): 163-1
6
8
.
[24]
Che
n
T
,
Popova E. Mainten
ance p
o
lic
ies
w
i
t
h
t
w
o
-
dim
e
nsio
nal
w
a
rran
t
y
.
Reli
abi
lity Engi
neer
in
g &
System
Safety
.
2002; 7
7
(1): 6
1
-69.
[25]
Gert
sbakh IB,
Kordo
n
sk
y
KB.
Paralle
l time scales an
d t
w
o
-
dimens
io
nal
manufactur
e
r a
nd ind
i
vid
u
a
l
customer w
a
rr
anties.
IIE transactions
. 19
98;
30(12): 1
1
81-1
189.
[26]
Do
ye
n L, Gau
doi
n O. Classe
s of
imperfect repa
ir
mod
e
ls base
d
on
r
e
d
u
c
tion of fai
l
ure
intens
it
y
or
virtual a
ge.
R
e
li
a
b
i
l
i
t
y En
gi
ne
eri
n
g
& System
Sa
fe
ty
. 2004; 8
4
(1): 45-5
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.