Sie sind auf Seite 1von 11

ARTICLE IN PRESS

Int. J. Production Economics 111 (2008) 686696


www.elsevier.com/locate/ijpe

Optimal buffer inventory and opportunistic preventive


maintenance under random production capacity availability
Romulo I. Zequeiraa,, Jose E. Valdesb, Christophe Berenguerc
College of Management of Technology (CDM), Chair of Technology and Operations Management (TOM), Ecole polytechnique federale de
a

Lausanne (EPFL), EPFL/CDM/ILEMT/TOM Odyssea, Station 5, CH-1015 Lausanne, Switzerland


b
Facultad de Matematica y Computacion, Universidad de La Habana, C. San Lazaro y L, CP 10400, La Habana, Cuba
c
CNRS - Institut Charles Delaunay/Systems Modelling and Reliability Laboratory, Universite de technologie de Troyes,
BP 2060 - 12, rue Marie Curie F-10010 Troyes cedex, France
Received 11 July 2005; accepted 8 February 2007
Available online 24 April 2007

Abstract

In this paper we study the determination of the optimal maintenance policy for a manufacturing facility and the optimal
buffer inventory to satisfy the demand during the interruption period due to a maintenance action. We consider the
possibility of imperfect production and that opportunities for the fabrication of the buffer inventory and opportunities to
carry out a maintenance action to the production facility are random.
r 2007 Elsevier B.V. All rights reserved.

Keywords: Preventive maintenance; Imperfect production; Buffer inventory; Opportunistic maintenance; Random production capacity
availability

1. Introduction consumer dissatisfaction due to defective items or


increased lead times.
Normal deterioration or unexpected external The probability of defective production or break-
shocks can provoke machine breakdowns or defec- downs can be reduced by carrying out preventive
tive production. Machine breakdowns are impor- maintenance. Preventive maintenance can also
tant causes of variability increase in process times improve production quality and yield (Sloan,
and ows of production systems, leading to reduced 2004). Nevertheless, preventive maintenance actions
manufacturing performance (Hopp and Spearman, may render unavailable the production facility with
2000). Imperfect production may increase total the consequent economic loss. If we are dealing with
production costs due to possible rework or material the nal products, unavailability of the manufactur-
loss. Imperfect production can also provoke ing facility may cause lost sales or increased costs
due to backordering. If we are considering produc-
tion processes, unavailability of one machine may
Corresponding author. provoke the disruption of the entire production
system.
E-mail addresses: romulo.zequeira@ep.ch (R.I. Zequeira),
vcastro@matcom.uh.cu (J.E. Valdes), One way to avoid economic losses due to
christophe.berenguer@utt.fr (C. Berenguer). production unavailability caused by preventive

0925-5273/$ - see front matter r 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.ijpe.2007.02.037
ARTICLE IN PRESS
R.I. Zequeira et al. / Int. J. Production Economics 111 (2008) 686696 687

maintenance actions is building a buffer inventory after the last preventive maintenance action for
to satisfy the demand during the production using the extra production capacity the rst (ran-
unavailability period. Nevertheless, inventory nan- dom) time it is available, and the optimal size S of
cing often incurs considerable opportunity costs. the buffer inventory.
Holding products in inventory also implies costs Note that the expected holding costs increase
due to allocation space, equipment required to when T decreases and, if the maintenance action is
support and move it and deterioration. Hence, made after the buffer inventory is built, as we
keeping minimum inventory levels is of practical assume in this paper, the expected number of
interest. Besides, it has been shown that process defective items increases when T increases. Simi-
reliability can be improved keeping low inventory larly, the expected shortage increases when S
levels together with increasing the skill levels of decreases and the expected holding costs increase
workers (Alles et al., 2000). Keeping low inventories when S increases. Therefore, the optimal determina-
is encouraged by modern management approaches tion of the values of T and S has a signicant
like just-in-time (Hopp and Spearman, 2000). practical interest.
Investment in capacity expansion remains one of
the most critical decisions for manufacturing 1.1. Literature review
organizations (Chou et al., 2007; Julka et al.,
2007). Because of high utilization of installed The analysis of breakdowns in production sys-
production capacities it can be expected that the tems has received attention in the past. Groenevelt
extra production capacity to build the buffer et al. (1992) study economic lot sizing decisions
inventory is available only at random times. considering the effects of machine breakdowns
Furthermore, it can be assumed that once it is and corrective maintenance. Analytical models of
available it can be used to produce the whole buffer. preventive maintenance and safety stock strategies
One example is when the extra production capacity are presented by Cheung and Hausman (1997)
is a facility that produces different kinds of items and Dohi et al. (2001) considering a production
under a non-preemptive policy. If the inter-arrival environment subject to random machine break-
times and quantities to be produced are random downs.
then the times at which this extra production Models with a practical appeal to study imperfect
capacity is available are random (Anupindi and production process have been proposed. Porteus
Tayur, 1998). Non-preemptive scheduling policies (1986) assumed the production process to be
are usual in manufacturing environments (Nahmias, perfectly functioning at the beginning of a lot.
2004). Production may shift to an out-of-control state each
In practice, unexpected production demand or time a unit is produced. Once the process is in the
spare parts shortage can cause the rescheduling of out-of-control state all items produced are defective.
the times of planned preventive maintenance. In Rosenblatt and Lee (1986) assumed that the
these cases a natural assumption is to consider that production system is functioning perfectly at the
periods between opportunities for the maintenance beginning of a lot and that the process may shift out
action to the production facility are random. of control at some random time and begin to
Motivated by the previous random characteristics produce a proportion q of defective items.
that are relevant to the management of production Models considering preventive maintenance of
facilities, we study in this paper a production system systems with buffers have been developed. van der
that can shift out of control and begin to produce a Duyn Schouten and Vanneste (1995) examined
proportion of defective items. Preventive mainte- maintenance optimization for production systems
nance actions to restore the production facility to a considering buffer capacity. In their model the
good-as-new condition have random durations and decision to start a preventive maintenance action
are opportunistic, to decrease overall costs. Besides, depends on the buffer level and on the condition of
preventive maintenance actions provoke the un- the installation. Meller and Kim (1996) studied a
availability of the production facility. To satisfy the two-machine system with a xed capacity buffer
demand during the unavailability period of the between the machines. Their model can be used to
manufacturing facility a buffer is built in advance determine the optimal buffer inventory that triggers
when extra production capacity is available. We are preventive maintenance on the rst production
interested in nding the optimal time threshold T machine. Perry and Posner (2000) studied a single
ARTICLE IN PRESS
688 R.I. Zequeira et al. / Int. J. Production Economics 111 (2008) 686696

machine that produces items to inventory. When a 1.3. Notation


machine breakdown occurs a repair operation
The following terms will have a specic meaning
begins. Otherwise a preventive maintenance action
in this paper:
is initiated after a xed time of continuous opera-
tion. Chelbi and Ait-Kadi (2004) presented a model
 Normal production corresponds to the produc-
to nd the optimal size of the buffer and the optimal
tion of the facility to which the maintenance
preventive maintenance time considering random
action is to be made.
duration for corrective and preventive maintenance
 Extra production corresponds to the production
actions. The buffer ensures the supply during the
of the facility that fabricates the buffer inven-
repair or the preventive maintenance. Rezg et al.
tory.
(2005) consider a production unit submitted to an
age-based maintenance policy. A buffer stock is
A cycle is dened as the period of time between two
built to guarantee the supply to the next stage in the
subsequent resumptions of fabrication in the
production system. They evaluate the average cost
normal production facility.
per time unit of the proposed strategy. Chelbi and
Rezg (2006) consider the optimization of a produc- a production rate of the extra
tion system under an age-based policy. They production facility,
optimize under an unavailability constraint the b consumption rate,
maintenance policy and the buffer that guarantees Cm expected cost of a maintenance action,
continuous supply of the production facility. T threshold time after a production
Salameh and Ghattas (2001) proposed a model to resumption: in a cycle, the fabrication
calculate the just-in-time buffer inventory that of the buffer inventory begins at the
minimizes the cost rate considering the sum of rst time after T at which the extra
the buffer inventory holding costs and the cost due production capacity is available,
to shortage in a cycle. Zequeira et al. (2004) S buffer inventory level,
extended the model of Salameh and Ghattas BS expected shortage cost in a cycle as a
(2001) considering the possibility of imperfect function of S,
production and that the duration and cost of the CT; S expected cost in a cycle as a function of
maintenance action depend on the state of the T and S,
production facility. CRT; S cost rate as a function of T and S,
DT; S expected cost in a cycle due to defective
items, as a function of T and S,
1.2. Main contribution d cost per defective item,
h inventory holding cost in $ per unit per
In this paper we propose a new model that time unit,
extends the model presented by Salameh and HT; S expected inventory holding cost in a
Ghattas (2001) by considering several aspects of cycle as a function of T and S,
the operation of production systems whose behavior LT; S expected cycle length as a function of T
is random, namely the time at which a production and S,
capacity is available to produce a given type of item, q probability that an item produced
and the opportunities to carry out preventive when the manufacturing facility is in
maintenance actions. That is, we consider that the the out-of-control state is defective,
extra production capacity to build a buffer inven- W random variable, in the cycle it is the
tory is available at random times and that the time from the instant in which the
maintenance action begins at the rst opportunity buffer fabrication is nished to the
for maintenance after the end of the fabrication of next maintenance opportunity time,
the buffer inventory. Further we consider in the Z parameter of the (exponential)
model the possibility of imperfect production. distribution of W,
A similar problem, usually known as opportunity- X random variable, in the cycle it is the
based age replacement, has been examined pre- time to shift to the out-of-control state,
viously to establish optimal replacement policies F x distribution function of X,
(Dekker and Dijkstra, 1992; Jhang and Sheu, 1999). f x probability density function of X,
ARTICLE IN PRESS
R.I. Zequeira et al. / Int. J. Production Economics 111 (2008) 686696 689

l failure rate of X when F x 1  elx , resume production until depletion of this


pT probability of imperfect production in remaining buffer inventory.
a cycle, 10 During the normal production period the
pS probability of shortage in a cycle, production rate is equal to the consumption
r shortage cost in $ per unit short, rate. Deliveries are made in small lots and
Y random variable, time after T at which therefore inventory levels are kept low.
the extra production capacity is
available, As we commented in the Introduction, one
u parameter of the (exponential) example in which assumptions 2 and 3 apply is
distribution of Y, when two types of production facilities for produ-
Z random variable, maintenance action cing the same good are available: one that can
duration, produce several types of products (the extra
EZ expected value of a random variable Z, production capacity, for example a exible manu-
Gz distribution function of Z, facturing system), and the normal production
gz probability density function of Z. facility, which produces only one product. Note
that under this setting assumption 3 is realistic
For a distribution function, say V x, its survival because if setup costs are high, then stopping the
function 1  V x will be denoted by V x. production of the buffer inventory and resuming
Values of random variables are denoted by the later would increase costs. Therefore, it can be
corresponding lowercase letters. considered that once the extra production facility
begins to build the buffer, then it will produce it
2. The model completely, and other products demand for the
extra production capacity will be satised later.
Assumptions: Finally, when the buffer inventory is much smaller
than the normal production batch, then producing
1. Maintenance actions return the facility to a it will likely be fast, reinforcing the practicality of
good-as-new condition. assumption 3.
2. Periods between subsequent times at which Note from assumption 7 that the shortage in a
the extra production capacity becomes cycle because of defective items cannot be lled up
available are exponentially distributed with from the buffer inventory of the cycle. This last
mean u1 . assumption corresponds in practice, for example, to
3 Once the extra production capacity is available the important case of failure modes not detectable
it can be used to produce a buffer inventory of during the production process, but after some time
size S, regardless of the value of S. of use by the nal consumers.
4. The time between opportunities for maintenance From assumption 8 we could wonder why not to
is exponentially distributed with mean Z1 . use the extra production capacity for normal
5. The normal production facility shifts out of production, if the reliability of the extra production
control at the random time X after a production capacity is very high. A case when it is not possible
resumption, and begins to produce defective to use the extra production capacity for the normal
items. Once the facility is in the out-of-control production is the example we have discussed
state each item produced is defective with previously, i.e. when the extra production capacity
probability q. has to be used for producing different items that
6. The out-of-control state of the normal produc- compete for that production capacity. Assumption
tion machine is not detectable before the 8 is acceptable for example when the buffer
preventive maintenance action. inventory is much smaller than the quantity
7. The defective state of items is not detected produced between preventive maintenance actions
immediately, but at a time much greater than the by the normal capacity. In that case the expected
cycle time. quantity of defective items produced by the extra
8. The probability of defective production for the production capacity is reduced because the total
extra production capacity is negligible. quantity produced is small.
9 If at the end of a maintenance action there is Assumption 9 can be justied by cost arguments.
buffer inventory available, the facility does not Maintaining the remaining buffer after a preventive
ARTICLE IN PRESS
690 R.I. Zequeira et al. / Int. J. Production Economics 111 (2008) 686696

maintenance actions incurs costs. Besides, using the The cost due to defective items dy; w, condi-
remaining buffer delays the resumption of the tioned on Y and W, is
normal production, with the consequent production Z TyS=aw
cost reduction. dy; w dqb F x dx.
0

2.1. Maintenance policy Hence the expected cost due to defective items is
Z 1 Z 1 Z TyS=aw
The policy we will consider is the following. Let y DT; S dqb ZuF x
0 0 0
be the rst time after the threshold time T, measured
euyZw dx dy dw.
from T, at which the extra production capacity is
available. After a production resumption a buffer The length of the cycle conditioned on Y, W and
inventory of size S begins to be built at time T y. Z, ly; w; z, is obtained as
Once the buffer inventory is built the maintenance 8
>
> S S S
action to the normal production facility begins at >T y w
< if zp ;
a b b
the rst opportunity for maintenance, at time ly; w; z
>
> S S
T y S=a w. See Fig. 1 where a realization of : T y a w z if z4 b :
>
the buffer inventory level as a function of time is
presented. We look for the values of T and S for Hence the expected length of the cycle is
which the cost rate is minimum. Z 1
1 1 S S
LT; S T Gz dz. (1)
u Z a b S=b
2.2. Cost rate function
The expected cost in a cycle is given by the sum of
When normal production and consumption rates the expected cost due to defective items, the
are equal and deliveries are made in small lots inventory holding cost, the expected shortage cost
(assumption 10) normal production inventory hold- and the preventive maintenance cost.
ing costs are kept at low levels. Therefore, we will Z 1 Z 1 Z TyS=aw
not consider those costs in the analysis but the CT; S dqb uZ
buffer inventory holding cost. Note that deliveries 0 0 0
uyZw
in small lots are encouraged by the just-in-time e F x dx dy dw
 
approach. S S2 S2
The expected buffer inventory holding cost in a h
2a 2b Z
cycle and the expected shortage cost in a cycle are Z 1
given by the following equations: rb Gz dz C m .
S=b
 2 
S S2 S
HT; S h , We look for the values of T and S for which the
2a 2b Z cost rate
Z 1 CT; S
CRT; S
BS rb Gz dz. LT; S
S=b
attains its minimum value, where LT; S is given by
Eq. (1).
It is natural to consider times to shift to imperfect
Extra production
S capacity availability production with increasing failure rates for produc-

tion facilities that deteriorate with the time of

operation. Nevertheless, in some cases it can be
assumed that the shift to imperfect production is
provoked by external actions (shocks), for example,
T y S/ w S/ time erroneous operator actions or abnormal environ-
z z mental conditions that can be assumed to have a
Preventive
Maintenance
Preventive constant arrival rate. That is, in these cases the
Maintenance
exponential distribution is a good model for the
Fig. 1. Realization of inventory level as a function of time. time to shift to imperfect production. Even when the
ARTICLE IN PRESS
R.I. Zequeira et al. / Int. J. Production Economics 111 (2008) 686696 691

failure rate is increasing the exponential distribution the following equation:


may be used under certain conditions as a good DS T; S
approximation for the time to shift to imperfect ( Z 1 !
production (Xie et al., 2000). One advantage of the 1 S 1
dqb Gz dz
use of the exponential distribution is its analytical a b S=b al
  
tractability. When the time to shift to imperfect 1 1 1
production is exponential, i.e. F x 1  elx , the  elTS=a
a 1 l=u 1 l=Z
expected cost due to defective production is given by Z 1 !
   1 1 S S 1
S 1 1 1 1  T Gz dz
DT; S dqb T u Z a b S=b l
a u Z l 1 l=u   "
   S 1 1 1 S 1
1 1 G T 
 elTS=a  . b b u Z a l
1 l=Z l
    #)
1 1 lTS=a
2.3. Determination of the optimal policy  1 e .
1 l=u 1 l=Z

The necessary condition to obtain the optimal It seems difcult to prove the convexity of the
value of S is given by cost rate function with respect to S. Nevertheless, it
U S S 0, (2) can be seen that under some conditions there exists,
for a value of T given and the other parameters
where xed, a unique value of S for which the cost rate
q q function attains its global minimum. Note that if the
U S S CT; S  LT; S  CT; S  LT; S following condition holds
qS qS
q2 q2
is given by the following equation: CT; S  LT; S  CT; S  LT; S40
(  qS 2 qS 2
S S 1 (3)
U S S DS T; S h LT; S
a b Z
then U S S is strictly increasing and if there is, for T
   )
S S 1 1 1 S xed, a positive value S of S for which U S S  0,
S G then the cost rate attains its global value at S  for
2a 2b Z a b b
(    the value of T given. Otherwise, if condition (3)
S 1 1 S S holds and for all S40 it holds that U S S40, then
r G T
b u Z a b the cost rate attains its global minimum value at
 Z 1 ) S 0, for the given value of T. It is obtained that
b q2 q2
1 Gz dz CT; S  LT; S  CT; S  2 LT; S
a S=b 2
   qS " qS
1 1 S Z 1Z 1  
 Cm G , Zwuy b S
a b b dq Zue f T y w
0 0 a2 a
where   Z TyS=aw #
1 S
DS T; S LT; S  g F x dx dy dw
Z Z b b 0
1 1
 
dqb ZueZwuy 1 1
0 0 h LT; S
(   a b
1 S    
 F T y w  LT; S 1 S2 S2 S S
a a  2 g
   Z TySw ) b 2a 2b Z b
  
1 1 S a
1 S 1 1 S S
 G F x dx dy dw. r g T
a b b 0 b b u Z a b
 
If the time to shift to an out-of-control state is 1 S
 Cm 2 g . 4
exponential with mean l1 then DS T; S is given by b b
ARTICLE IN PRESS
692 R.I. Zequeira et al. / Int. J. Production Economics 111 (2008) 686696

Note that, in our setting, when z ! 1 we should where


have gz ! 0. That is, the probability that a
q q
maintenance action duration is too large is negli- U T T CT; S  LT; S  CT; S  LT; S
qT qT
gible. Hence from (4) it follows that for values of S
large enough it can be expected that condition (3) is given by the following equation:
holds. Therefore, if costs due to imperfect produc-  2 
tion and holding inventory are greater than zero S S2 S
U T T DT T; S  h
and S is large enough then U S S is strictly 2a 2b Z
Z 1
increasing in S.
We give now a sufcient conditionEq. (6)for  rb Gz dz  C m ,
S=b
(3) to hold, which is independent of the distribution
of the duration of the maintenance action. Eq. (4) where
can be rewritten as Z 1 Z 1
DT T; S dqb ZueZwuy
q2 q2 0 0
CT; S  LT; S  CT; S  2 LT; S "  
2
qS Z 1Z 1 qS S
b  F T y w  LT; S
dq ZueZwuy 2 a
0 0 a Z TyS=aw #
 
S  F x dx dy dw.
f T y w dy dw
a 0
 
1 1 When the time to shift to an imperfect production
LT; S h LT; S
a b is exponential, DT T; S is given by the following
   2  equation:
1 S S S2 S
 2g DT; S h
b b 2a 2b Z ( Z !
   S 1
1
1 1 S S DT T; S dqb Gz dz
rb T Cm . 5 b
S=b l
u Z a b
    
Therefore, if 1 1 lTS=a
 1 e
1 l=u 1 l=Z
rXBr T; S, (6)   
1 1
where Br T; S is given by  elTS=a
1 l=u 1 l=Z
1 1  )
Br T; S 1 1 S
b T 1=u 1=Z S=a S=b  T .
u Z a
  2  
S S2 S
 DT; S h Cm ,
2a 2b Z It is obtained that
then U S S is strictly increasing.
From condition (6) it can be expected that U S S q2 q2
2
CT; S  LT; S  CT; S  2 LT; S
is increasing for high values of the shortage costs r. qT Z Z qT
1 1
It is not difcult to nd a numerical example to dqbLT; S ZueZwuy
show that the cost rate function CRT; S is not 0 0
 
convex as a function of T for S and the other S
parameters xed. Nevertheless, we will show that f T y w dy dw.
a
for S and the other parameters xed there exist,
under some conditions, a unique value of T for Hence if for all T
which the cost rate function CRT; S attains a
Z 1Z 1
global minimum.
dqbLT; S ZueZwuy
The necessary condition for the optimal T is given 0 0
by  
S
f T y w dy dw40, 8
U T T 0, (7) a
ARTICLE IN PRESS
R.I. Zequeira et al. / Int. J. Production Economics 111 (2008) 686696 693

i.e. if the expected costs due to imperfect production time equal to 14 days between occurrences of
are greater than zero and if there exists, for S and opportunities to begin building the buffer inventory,
the other parameters xed, a positive value T  of T and to opportunities for the maintenance action
for which condition (7) holds, then the cost rate occurring each 2 days, in average. We will consider
function CRT; S attains its global minimum at T  . an exponential time to shift to an out-of-control
Otherwise if condition (8) holds and for all T40 it state with failure rate l 1 year1 .
holds that U T T40 then the optimal value of T is We will consider that the duration of the
0. Note that under the assumptions of our model if maintenance action has an exponential distribu-
there is no imperfect production then it is mean- tion. Although other distributions like the Lognor-
ingless to set a nite value for T and the optimal mal are suitable to model time to repair since
value for T is T 1, i.e. not to carry out they do not have the memoryless property, the
maintenance actions. exponential distribution is more tractable analyti-
Hence if conditions (3) and (8) hold then the cally and in a number of situations it is a good
optimal values of T and S can be obtained using the model for the time to repair since it gives greater
following algorithm: probability to short durations of the maintenance
action.
1. Set an initial value S 1 of S. Fig. 2 presents the cost rate CRT; S as a
2. Repeat until no changes in S j and T j occur, function of S and T for values of S between 0 and
j 1; 2; . . . . 379 units and for T between 4 and 82 days. Values 0
2.1. Set S Sj . Let T j be the value of T for which and 379 units for buffer inventory level S corre-
Eq. (7) holds. If T j o0 then set T j 0. spond to a probability of 1 and 0.01, respectively, of
2.2 Set T T j . Let Sj be the value of S for which shortage in a cycle. Values 4 and 81 days for T
Eq. (2) holds. If S j o0 then set S j 0. correspond to a probability of 0.01 and 0.20,
respectively, that before time T the production
In practice, we can set an interval for T and an shifts to imperfect production. Note from Fig. 2
interval for S considering reasonable values for that a global minimum exists for the cost rate
the probability of imperfect production and the function CT; S for the interval considered for T
shortage probability. Let pT and be pS be values and S. Besides, for this data set the cost rate
of the probability of imperfect production before function CRT; S is rather sensitive to both T
T and the shortage probability in a cycle, respec- and S.
tively. An interval for T inside which the optimal To look for the optimal values analytically we set
value of T can be expected to be is 0; T max , pT 0:9999 and pS 0:0001. In correspondence
where T max is the value of T for which pT is close to with these probabilities, the intervals we set for the
one. Similarly, an interval for S inside which the search are [0 year, 9.2 year] and [0 units, 758 units]
optimal value of S can be expected to be is 0; Smax , for T and S, respectively. For the data set and
where S max is the value of S for which pS is close to the intervals for T and S we are considering it can
zero. be checked that condition (6) holds. Hence there is,
Note that for T and S given the costs due to for each value of T, a unique value of S for which
holding inventory and to defective production the cost rate attains its global minimum. Besides,
increase when the buffer replenishment rate a since the expected costs due to imperfect production
decreases. In the Numerical example section are greater than zero, then condition (8) holds.
we study the dependence of the optimal T on S Therefore, there is, for each value of S, a unique
for different values of the buffer replenishment value of T for which the cost rate attains its
rate a. global minimum. The optimal values of T and S
are found to be 0.072 year, that is 26.3 days, and 98
3. Numerical example units, respectively. The minimum cost rate is
$3443.9/year.
Consider the following data: a 15 000 units per If the buffer inventory is not built, i.e. if S 0,
year, b 30 000 units per year, u1 14 days, then the optimal value of T is 0.089 years, which
Z1 2 days, EZ 1 day, q 0:1, d $10 per unit, corresponds to 32.5 days. This leads to a cost rate of
C m $150, h $25 per unit per year and r $1 $3723.2/year, which represents an increase of 8.1%
per unit. Note that these data correspond to a mean in relation to the minimum attainable cost rate
ARTICLE IN PRESS
694 R.I. Zequeira et al. / Int. J. Production Economics 111 (2008) 686696

5500

Cost rate ($/year) 5000

4500

4000

3500

3000
100
80 400
60 300
T( 40
da 200
ys) 20 100
its)
0 0 S (un

Fig. 2. Cost rate as a function of T and S.

80

70

60
Optimal T (days)

50

40

30

20

10
0 100 200 300 400 500 600 700 800
S (units)

Fig. 3. Optimal value of T as a function of S for a 3000 units per year (dotted line), a 6000 units per year (dashed line) and a 15 000
units per year (solid line).

when S and T are set at optimal values. If we The optimal value of T as a function of S for a
set T 0 then we obtain that the optimal 3000 units per year (dotted line), a 6000 units per
buffer inventory is S 147 for which the cost year (dashed line) and a 15 000 units per year (solid
rate is $4646.4/year. This corresponds to an in- line) is represented in Fig. 3. Note that for values
crease of 34.9% of the cost rate in comparison relatively small of S we have that the optimal T
with the attainable minimal cost rate if both increases when a increases. This can be explained in
decision variables T and S are set to the optimal the following way. For small values of S the buffer
values. inventory holding cost in a cycle is relatively small.
ARTICLE IN PRESS
R.I. Zequeira et al. / Int. J. Production Economics 111 (2008) 686696 695

12000

11000

10000
Minimum cost rate ($/year)
9000

8000

7000

6000

5000

4000

3000
0 100 200 300 400 500 600 700 800
S (units)

Fig. 4. Minimum cost rate as a function of S for a 3000 units per year (dotted line), a 6000 units per year (dashed line) and a 15 000
units per year (solid line).

For the values of T and S given, the expected cost due References
to defective production in a cycle decreases when a
increases due to a faster buffer fabrication. Hence Alles, M., Amershi, A., Datar, S., Sarkar, R., 2000. Information
when S has a value relatively small we can set and incentive effects of inventory in JIT production. Manage-
ment Science 46, 15281544.
relatively high values for T, still maintaining the cost Anupindi, R., Tayur, S., 1998. Managing stochastic multiproduct
due to defective production in a cycle at reasonable systems: model, measures and analysis. Operations Research
levels. As a consequence, when S has a value relatively 46, 98111.
small the optimal T increases when a increases. Chelbi, A., Ait-Kadi, D., 2004. Analysis of a production/
For sufciently high values of S the inventory inventory system with randomly failing production unit
submitted to regular preventive maintenance. European
holding costs behave as a quadratic function of S,
Journal of Operational Research 156, 712718.
and the cost due to defective production tends to be Chelbi, A., Rezg, N., 2006. Analysis of a production/inventory
linear as a function of S. Therefore, when S is set to system with randomly failing production unit subjected to a
a value relatively high and T is xed, the inventory minimum required availability level. International Journal of
holding cost in a cycle has a high value that Production Economics 99, 131143.
Cheung, K.L., Hausman, W.H., 1997. Joint determination of
increases when a decreases and therefore outweighs
preventive maintenance and safety stocks in an unreliable
defective production costs in a cycle. Hence for high production environment. Naval Research Logistics 44,
values of S we will carry out the maintenance action 257272.
less frequently when a decreases to keep the Chou, Y.C., Cheng, C.T., Yang, F.C., Liang, Y.Y., 2007.
inventory holding cost contribution to the cost rate Evaluating alternative capacity strategies in semiconductor
at reasonable levels. This corresponds to the results manufacturing under uncertain demand and price
scenarios. International Journal of Production Economics
represented in Fig. 3, where for high values of S the 105, 591606.
optimal value of T increases when a decreases. Dekker, R., Dijkstra, M.C., 1992. Opportunity-based age
The minimum cost rate as a function of S, i.e. replacement: Exponentially distributed times between oppor-
taking the optimal value of T for each value of S, is tunities. Naval Research Logistics 39, 175190.
represented in Fig. 4 for a 3000 units per year Dohi, T., Okamura, H., Osaki, S., 2001. Optimal control of
preventive maintenance schedule and safety stocks in an
(dotted line), a 6000 units per year (dashed line)
unreliable manufacturing environment. International Journal
and a 15 000 units per year (solid line). Note that of Production Economics 74, 147155.
the minimum cost rate increases and the optimal S van der Duyn Schouten, F.A., Vanneste, S.G., 1995. Main-
decreases when a decreases, as it can be expected. tenance optimization of a production system with buffer
ARTICLE IN PRESS
696 R.I. Zequeira et al. / Int. J. Production Economics 111 (2008) 686696

capacity. European Journal of Operational Research 82, Porteus, E.L., 1986. Optimal lot-sizing, process quality improve-
323338. ment and setup cost reduction. Operations Research 34,
Groenevelt, H., Pintelon, L., Seidmann, A., 1992. Production lot 137144.
sizing with machine breakdowns. Management Science 38, Rezg, N., Chelbi, A., Xie, X.L., 2005. Modeling and optimizing a
104123. joint inventory control and preventive maintenance strategy
Hopp, W.J., Spearman, M.L., 2000. Factory Physics: Founda- for a randomly failing production unit: Analytical and
tions of Manufacturing Management. McGraw-Hill Higher simulation approaches. International Journal of Computer
Education, New York. Integrated Manufacturing 18, 225235.
Jhang, J.P., Sheu, S.H., 1999. Opportunity-based age replace- Rosenblatt, M.J., Lee, H.L., 1986. Economic production cycles with
ment policy with minimal repair. Reliability Engineering & imperfect production processes. IIE Transactions 18, 4855.
System Safety 64 (3), 339344. Salameh, M.K., Ghattas, R.E., 2001. Optimal just-in-time buffer
Julka, N., Baines, T., Tjahjono, B., Lendermann, P., Vitanov, V., inventory for regular preventive maintenance. International
2007. A review of multi-factor capacity expansion models for Journal of Production Economics 74, 157161.
manufacturing plants: Searching for a holistic decision Sloan, T.W., 2004. A periodic review production and main-
aid. International Journal of Production Economics 106, tenance model with random demand, deteriorating equip-
607621. ment, and binomial yield. Journal of the Operational
Meller, R.S., Kim, D.S., 1996. The impact of preventive Research Society 55, 647656.
maintenance on system cost and buffer size. European Xie, M., Kong, H., Goh, T.N., 2000. Exponential approximation
Journal of Operational Research 95, 577591. for maintained Weibull distributed component. Journal of
Nahmias, S., 2004. Production and Operations Analysis, fth ed. Quality in Maintenance Engineering 6, 260269.
Irwin/McGraw-Hill, New York. Zequeira, R.I., Prida, B., Valdes, J.E., 2004. Optimal buffer
Perry, D., Posner, M.J.M., 2000. A correlated M/G/1-type queue inventory and preventive maintenance for an imperfect
with randomized server repair and maintenance modes. production process. International Journal of Production
Operations Research Letters 26, 137147. Research 42, 959974.

Das könnte Ihnen auch gefallen