Professional Documents
Culture Documents
, Hennie Husniah
,
and Bermawi P. Iskandar
_
(2)
where F
0
(x) is the base failure distribution function.
Failures over time follow a non-homogeneous Poisson
process with intensity function
(x|y) = y
0
_
xy
1
_
(3)
where
0
(x) is the base intensity function.
2.2. Modelling First Failure
For a repairable product sold with a two-dimensional
warranty, one needs to model the products degradation
which takes into account both age and usage. [8] have in-
troduced a more appropriate model which uses the accel-
erated failure time (AFT) model, to represent the effect
of usage rate on degradation. Let y
0
denotes the nominal
usage rate value associated with component reliability.
When the actual usage rate is different from this nominal
value, the component reliability can be affected and this
in turn affects the product reliability. As the usage rate
increases above the nominal value, the rate of degrada-
tion increases and this, in turn, accelerates the time to
failure. Consequently, the product reliability decreases
[increases] as the usage rate increases [decreases]. Us-
ing the AFT formulation, if X
0
denotes the time to rst
failure under usage rate y
0
then we have
X
X
0
=
_
y
0
y
_
(4)
Furthermore, if the distribution function for X
0
is given
by F
0
(x;
0
) where
0
is the scale parameter, then the
distribution function for X is the same as that for X
0
but
with a scale parameter given by
(y) =
_
y
0
y
_
0
(5)
with 1. Hence, we have
F (x; (y)) = F
0
_
(y
0
/y)
x;
0
_
(6)
The hazard and the cumulative hazard functions associ-
ated with F (x, (y)) are given by
h(x, (y)) = f (x, (y))/(1F(x; (y)) (7)
and
H(x, (y)) =
_
x
0
h
_
x
, (y)
_
dx
(8)
respectively, where f (x, (y)) is the associated density
function.
2.3. Warranty Policy
The product is sold with a two-dimensional non-
renewing failure replacement warranty (NFRW) together
with a rectangle warranty region
W
= [0,W) [0, U ).
Here W is the warranty expiry time and U is the usage
limit. With NFRW, all failures under warranty are recti-
ed at no cost to the buyer. It is assumed that the recti-
cation is done through a minimal repair and the repaired
product comes with the original warranty. The warranty
38
Downloaded 04 Jun 2012 to 180.246.31.163. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/about/rights_permissions
FIGURE 1. a). Warranty region and b).maintenance region
ceases at the rst instance when the age of the product
reaches W or its usage reaches U , whichever occurs rst.
The two-dimensional warranty region is given by
W
y
=
_
W for y U /W,
U /y for y >U /W.
(9)
3. PERIODIC REPLACEMENT POLICY
y
FOR A PRODUCTS SOLD WITH
TWO-DIMENSIONAL WARRANTY
A periodic replacement policy for a given usage rate y
will be considered here and also the policy for various
values of usage rate. Let
y
denotes the parameter of a
periodic replacement policy for that rate. The periodic
replacement policy is dened as follows.
For a given usage rate y, the product is repaired with
minimal repair when it fails at age x, with x <
y
, and it
is replaced with a new one if its age reaches
y
. We seek
the optimal value of
y
which minimizes the long-run
average cost to the buyer. If
y
denotes the optimal
value then, as y varies, we have a set of points (
y
, y
y
)
dening a curve as indicated in Figure 2. Let denotes
a closed region bounded by this curve. The replacement
policy for a various values of y is dened as follows.
For a given usage rate y, the product is repaired with
minimal repair when it fails at age x, with x <
y
, and it
is replaced with a new one if its age reaches
y
.
Let
y
denotes the optimal value of
y
which mini-
mizes the expected cost time per unit to the buyer. The
problem is to nd a pair of points (
y
, y
y
) dening a
curve as indicated in Figure 2. Let denotes a closed re-
gion bounded by this curve. The replacement policy for
a various values of y can be seen as follows.
The product is repaired with minimal repair when
failure occurs in region and it is replaced with a new
one when its age at
y
. Furthermore, the expected cost
per unit time, denotes as J(
y
), is obtained as follows.
J (
y
) =
E[Cost per cycle]
E[Cycle length]
(10)
From above equation, the expected cycle length for
this replacement policy is
y
. Since all failures, either
occur within (0,W
y
) or (W
y
,
y
), are rectied with min-
imal repairs, then they will follow a non-homogeneous
Poisson process in the interval (0,
y
) with the inten-
sity function
y
(x) = h(x; (y)). For a product with two-
dimensional NFRW, the consecutive times between re-
placement form a failure process with a renewal cycle.
Product replacement may be carried out either after or
prior the warranty ceases, i.e. when
y
W
y
or
y
<W
y
.
Next the replacement time both for
y
W
y
and
y
<W
y
will be modelled.
Let C
r
, C
m
(with C
m
> C
r
) and C
d
the cost of each
repair, the cost of a replacement and the cost incurred
by a buyer for any minimal repair done respectively.
When
y
W
y
, the expected cost per unit time for a two-
dimensional warranted product is
C
d
_
W
y
0
y
(x)dx +C
r
+(C
d
+C
m
)
_
y
W
y
y
(x)dx (11)
And hence the expected cost per cycle is given by
J(
y
) =
C
d
R(W
y
) +C
r
+(C
d
+C
m
)[R(
y
) R(W
y
)]
y
(12)
with R(W
y
) =
_
W
y
0
y
(x)dx.
On the other hand, when
y
< W
y
the expected cost
per unit time for a two-dimensional warranted product is
given by
C
d
_
y
0
y
(x)dx +C
r
(13)
with the expected cost per unit time
J (
y
) =
C
d
_
y
0
y
(x)dx +C
r
y
(14)
39
Downloaded 04 Jun 2012 to 180.246.31.163. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/about/rights_permissions
In the case
y
= W
y
, the expected cost per unit time
in equations (12) is equivalent to that in (14). In other
words, the optimal solution of
y
time can be derived for
each case.
3.1. Local Optimal Solution
Here, for a warranted product, optimal strategies
y
are found by separating two cases: after the expire of the
warranty
y
W
y
and prior the expiry of the warranty
y
<W
y
.
The investigation for the local optimal solution will be
done by differentiating the cost function J
i
(
y
), i = 1, 2
where condition 1 and 2 represent
y
W
y
and
y
< W
y
and set the resulting derivative to zero. Following the
proof in [4]; Lemma 1, Theorems 1, and 2 are derived
for each condition.
Lemma 1. Let
y
(
y
) is an Increasing Failure Rate
(IFR), then L(
y
) is a non-negative increasing function
of
y
with lim
y
0
L(
y
) = 0 and lim
L(
y
) > 0.
Proof. Note that is
y
(
y
) IFR, then
y
(
y
) > 0. This
means that
dL(
y
)
d
y
= (
y
) +
y
y
(
y
)
y
(
y
) =
y
y
(
y
) > 0
Then we have
lim
y
0
L(
y
) = lim
y
0
L(0) = 0
and
lim
L(
y
) = lim
L() > 0
Theorem 1. Let
y
(
y
) is an IFR, is an IFR, then for
a product sold with two-dimensional NFRW policy, the
optimal replacement strategy after the expiry of the war-
ranty,
y
[W
y
, ) is given by:
1.
y
=, whenever L()
C
r
C
m
R(W
y
(C
m
+C
d
)
2. a nite and unique
y
> W
y
satisfying
L(
y
) =
C
r
C
m
R(W
y
(C
m
+C
d
)
with the expected cost per
unit is given by J(
y
) =
C
r
+C
m[R(
y
)R(W
y
)]+C
d
R(
y
)
y
whenever L(W
y
) <
C
r
C
m
R(W
y
(C
m
+C
d
)
< L().
Proof. If L()
C
r
C
m
R(W
y
(C
m
+C
d
)
then
dJ(
y
)
d
y
< 0. This means
that J(
y
) is a decreasing function with the respect to
y
in the interval [W
y
, ), and consequently
y
= . On
the other hand, if L(W
y
) <
C
r
C
m
R(W
y
(C
m
+C
d
)
< L(), then from
Lemma 1. L(
y
) is a non-negative increasing function
with the respect to
y
. Consequently, there exist a nite
and unique
y
, satisfying
dJ(
y
)
d
y
= 0, such that L(
y
) =
C
r
C
m
R(W
y
)
(C
m
+C
d
)
. The resulting expected cost per unit time is
given by J(
y
) =
C
r
+C
m[R(
y
)R(W
y
)]+C
d
R(
y
)
y
.
Theorem 2. Let
y
(
y
) is an IFR, then for a product
sold with two-dimensional NFRW warranty policy, the
optimal replacement strategy prior the expiry of the war-
ranty,
y
[0,W
y
) is given by:
1.
y
=W
y
, whenever L(W
y
)
C
r
C
d
2. a nite and unique
y
< W
y
satisfying L(
y
) =
C
r
C
d
with the expected cost per unit is given by J(
y
) =
C
d
R(
y
)+C
r
y
whenever L(W
y
) >
C
r
C
d
.
Proof. Since L(
y
) is a non-negative increasing function
in
y
then
dJ(
y
)
d
y
0 for
y
[0,W
y
) whenever L(W
y
)
C
r
C
d
. Hence, J(
y
) is a decreasing function in
y
within the
interval [0,W
y
), which means that
y
=W
y
with J(
y
) =
J(W
y
) =
C
d
R(W
y
+C
r
W
y
. On the other hand, if L(W
y
) >
C
r
C
d
,
then there exist a nite and unique
y
< W
y
, satisfying
dJ(
y
)
d
y
=0, such that L(
y
) =
C
r
C
d
. The associated expected
cost per unit time is given by J(
y
) =
C
d
R(
y
+C
r
y
.
3.2. Global Optimal Solutions
In practice, some experts obey the separation of re-
placement time in
y
W and
y
> W, and the appro-
priate strategy is a global optimal solution. If
G
y
is the
global optimal solution, with the expected cost per unit
time J(
G
y
), then it can be obtained by combining the
local optimal solution
y
in each case. The following
Corollary is a direct consequences of Theorems 1 and
2 related to
G
y
and J(
G
y
).
Corollary 1. Assume that a product sold with two-
dimensional NFRW warranty policy and the hazard rate
for the product, h
y
(x), which follows an IFR for x (0,
y
]
with C
r
, C
d
> 0, then the global optimal replacement so-
lution is given by:
1.
G
y
=
y
>W
y
with the expected cost per unit time
is given by (12) whenever L(W
y
) <
C
r
C
m
R(W
y
)
(C
m
+C
d
)
.
2.
G
y
=
y
W
y
with the expected cost per unit time is
given by (14) whenever
C
r
C
m
R(W
y
)
(C
m
+C
d
)
L(W
y
)
C
r
C
d
.
40
Downloaded 04 Jun 2012 to 180.246.31.163. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/about/rights_permissions
TABLE 1. The optimal replacement solution,
Gy
with C
r
= 5, C
d
=C
m
= 1,
0
= 1, y
0
= 2 for various usage rates y.
y
Gy
J(
Gy
) y
Gy
J(
Gy
) y
Gy
J(
Gy
)
0.60 9.518014110 1.027945524 1.5 3.442651863 2.904737510 8.0 0.2795084972 35.77708764
0.65 8.415852999 1.155601815 2.0 2.236067977 4.472135955 8.5 0.2552123033 39.18306395
0.70 7.503934828 1.286924823 2.5 1.600000000 6.250000000 9.0 0.2342427896 42.69074842
0.75 6.738501867 1.421402737 3.0 1.217161239 8.215838362 9.5 0.2159954425 46.29727314
0.80 6.087897831 1.558501845 3.5 0.965890576 10.35313962 10.0 0.2000000000 50.00000000
0.85 5.528712317 1.697660226 4.0 0.790569415 12.64911064 10.5 0.1858857282 53.79649152
0.90 5.043296764 1.838281670 4.5 0.662538660 15.09345885 11.0 0.1733568344 57.68448665
0.95 4.618118188 1.979729541 5.0 0.565685424 17.67766953 11.5 0.1621747493 61.66188043
1.00 4.242640687 2.121320344
y 1
Remark 1. If the cost of down time, C
d
, is relatively
smaller than the replacement cost C
r
, then the corollary
reveals that optimal replacement period would be longer
than the warranty period. In contrast, if that cost is
relatively greater than the cost of replacement, then it
is optimal to replace the product before the warranty
ceases. Logically, this is acceptable to avoid unnecessary
additional cost of down time. In this case, if L(W
y
) >
C
r
C
d
then an early replacement before the warranty expires
also optimal.
Remark 2. Beside L
y
(x), the quantity h
y
(x) also inu-
ences the optimal replacement period. This is due to the
monotonic property of h
y
(x), which in turn cause the
monotonic property of L
y
(x).
Remark 3. The model discussed in [4] is nested in the
current model, that is when y = y
0
. In this case, if x W
then (12) reduces to equation (4) in [4], meanwhile if
x <W, then (14) collapses to equation (6) in [4].
4. NUMERICAL EXAMPLE
A numerical example is presented to illustrate the prop-
erties of the optimal replacement discussed in the pre-
vious section. Assume that the time to the rst fail-
ure under usage rate, X
0
, follows a Weibull distribution
with the conditional distribution function F
0
(x;
0
) =
1 exp(x/
0
)
_
y
y
0
_
x
1
0
.
Moreover, since all repairs are minimal, then
y
(x) =
h(x; (y)) with cumulative hazard function R(W
y
) =
W
y
/[
0
(y
0
/y)
0
=1 (year) and =2; (3) AFT parameter: =1.5; (4)
costs: C
m
=1, C
r
=5, and C
d
=1. Table 1. shows the op-
timal solution
G
y
for warranty area
W
= [0, 2) [0, 2).
Figure 2 and 3 show the visual solutions of the optimal
replacement period
G
y
for various usage rates.
Tables 1 shows that
1. Numerical examples conform with the properties
predicted by Theorem 1 and 2 in Section 3.1 and the
corollary 1 in Section 3.2, e.g. the increase of usage
rate shortening the length of periodic replacement
interval
G
y
. This is plausible since a product with a
high usage rate would be replaced more often than
that with a lower usage rate.
2. The increase of down time cost C
d
widening the
length of periodic replacement interval
G
y
.
FIGURE 2. Maintenance region for y 1.
41
Downloaded 04 Jun 2012 to 180.246.31.163. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/about/rights_permissions
FIGURE 3. Maintenance region for y > 1.
3. The cost of down time, C
d
, is relatively smaller than
the replacement cost C
r
, then the corollary reveals
that optimal replacement period would be longer
than the warranty period (Remark 1). However, if
the cost of down time is relatively greater than the
cost of replacement, then it is optimal to replace the
product before the warranty ceases.
5. CONCLUSION
In this paper, a periodic replacement policy for a prod-
uct with a two-dimensional warranty has been studied
particularly for product with Non Renewing Warranty
using one dimensional approach. One can study other
two-dimensional replacement policies such an age re-
placement policy or an opportunity-based age replace-
ment [16] incorporates with a two-dimensional approach
where product failures is modelled by two-dimensional
random points occurring over the maintenance region
[11]. These topics are currently under investigation.
REFERENCES
1. S. C. Yang, and J. A. Nachlas, Bivariate Reliability and
Maintenance Planning Models, IEEE Transactions on
Reliability, 2001, Vol. 50, No.1, pp. 2635.
2. G. M. Jung, and D. H. Park, Optimal maintenance
policies during the post-warranty period, Reliability
Engineering and System Safety, 1999, 82, pp. 123185.
3. K. M. Jung, S. S. Han, and D. H. Park, Optimization
of cost and downtime for replacement model following
the expiration of warranty, Reliability Engineering and
System Safety, 2008, 93, pp. 9951003.
4. R. H. Yeh, M. Y. Chen, and C. Y. Lin, Optimal periodic
replacement policy for repairable products under free-
repair warranty, European Journal of Operational
Research, 2007, 176, pp. 16781686.
5. J. S. Dagpunar, Hybrid minimal repair and age
replacement maintenance policies with non-negligible
repair time, Naval Research Logistics, 1994, 41, pp.
10291037.
6. T. Nakagawa, Modied Periodic Replacement with
minimal Repair at failure, IEEE Transactions on
Reliability, 1981, 30, pp. 165168.
7. J. E. Goulionis, and V. K. Benos, A maintenance
policy under markovian deterioration, Advances and
Applications in Statistics, 2007, 7(3), pp. 357388.
8. J. Nat, B. P. Iskandar, and D. N. P. Murthy, A repair-
replace strategy based on usage rate for items sold with a
two-dimensional warranty, Reliability Engineering and
System Safety, 2008.
9. D. N. P. Murthy and R. J. Wilson, Modelling two-
dimensional warranties, in Proceedings of the Fifth
International Symposium on Applied Stochastic Models
and Data Analysis, Granada, Spain, 1991, pp. 481492.
10. B. P. Iskandar, R. J. Wilson, and D. N. P. Murthy,
A new repair-replace strategy for items sold with a
two-dimensional warranty, Computers & Operations
Research, 2005, 32, pp 669682.
11. D. N. P. Murthy, and W. R. Blischke, Warranty Cost
Analysis, Marcel Dekker Inc., New York,1993.
12. R. E. Barlow, and L. Hunter, Optimal preventive
maintenance policies, Operations Research, 1960, 8, pp.
90100.
13. J. Lawless, J. Hu, and J. Cao, Methods for estimation of
failure distributions and rates from automobile warranty
data, Lifetime Data Analysis, 1995, 1, pp. 227240.
14. J. Lawless, Statistical Models and Methods for Lifetime
Data, John Wiley, New York, 1982.
15. W. R. Blischke, and D. N. P. Murthy, Reliability:
Modeling,Prediction, and Optimisation, John Wiley, New
York, 2000.
16. B. P. Iskandar, and H. Sandoh, An opportunity-based age
replacement policy considering warranty, International
Journal of Reliability, Quality and Safety Engineering,
1999, 1(3), pp. 229236.
42
Downloaded 04 Jun 2012 to 180.246.31.163. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/about/rights_permissions