Two Echleon by JHa JK

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Two-echelon supply chain inventory model with controllable

lead time and service level constraint


J.K. Jha
*
, Kripa Shanker
Department of Industrial and Management Engineering, Indian Institute of Technology Kanpur, Kanpur, Uttar Pradesh 208 016, India
a r t i c l e i n f o
Article history:
Received 26 November 2008
Received in revised form 29 April 2009
Accepted 29 April 2009
Available online 6 May 2009
Keywords:
Vendorbuyer integrated system
Controllable lead time
Continuous review policy
Service level
a b s t r a c t
This paper considers a two-echelon supply chain inventory problem consisting of a single-vendor and a
single-buyer. In the system under study, a vendor produces a product in a batch production environment
and supplies it to a buyer facing a stochastic demand, which is assumed to be normally distributed. Also,
buyers lead time is controllable which can be shortened at an added cost and all shortages are backor-
dered. A model has been formulated for an integrated vendorbuyer problem to jointly determine the
optimal order quantity, lead time and the number of shipments from the vendor to the buyer during a
production cycle while minimizing the total expected cost of the vendorbuyer integrated system. It is
often difcult to estimate the shortage cost in inventory systems. Therefore, instead of having a shortage
cost term in the objective function, a service level constraint (SLC) is included in the model that requires a
certain proportion of demands to be met in each cycle. An efcient procedure has been suggested to nd
the bounds on number of shipments and then, an algorithm is developed to obtain the optimal solution of
the proposed model. A numerical example is included to illustrate the algorithmic procedure and the
effects of key parameters are studied to analyze the behavior of the model. Finally, the savings of buyer
and vendor are investigated from implementation of joint optimization model over the model in which
they minimize their own cost independently.
2009 Elsevier Ltd. All rights reserved.
1. Introduction
Nowadays, companies can no longer compete solely as individ-
ual entities in the constantly changing business world. Globaliza-
tion of market and increased competition force organizations to
rely on effective supply chains to improve their overall perfor-
mance. Successful supply chain management requires a change
from managing distinct function to integrating activities into key
supply chain processes. Integration between two different business
entities is an important way to gain competitive advantages as it
lowers the joint total cost of the system. Therefore, in modern
enterprises, the integration of vendorbuyer inventory system is
an important issue.
The joint optimization concept for buyer and vendor was initi-
ated by Goyal (1976). Subsequently, numerous scholars developed
integrated inventory models under various assumptions. For exam-
ple, Banerjee (1986) assumed that the vendor manufactures at a -
nite rate and considered a joint economic-lot-size model in which
a vendor produces to order for a buyer on a lot-for-lot basis. Goyal
(1988) relaxed the lot-for-lot policy and suggested that vendors
economic production quantity should be an integer multiple of
buyer purchase quantity. As a result of using the approachsuggested
in the Goyals (1988) model, signicant reduction in inventory cost
can be achieved. Later, Pan and Yang (2002) improved Goyals
(1988) model by considering lead time as a controllable factor in
the model and obtained a lower joint total expected cost and shorter
lead time. Goyal (2003) suggested a simple procedure for determin-
ingtheoptimal operatingpolicyfor thePanandYangs (2002) model.
Yang and Pans (2004) extended Pan and Yangs model (2002) by
incorporating quality related issue. Hoque and Goyal (2006) pro-
posed a generalized model for the same system as in Pan and Yang
(2002) by transferring the lot from the vendor to the buyer with
equal or unequal sizedbatches. Ouyang, WuandHo(2007) extended
YangandPans (2004) model byallowingfor shortages andusingthe
reorder point as a decision variable. Several researchers have shown
that in integrated models, one partners gain exceeds the other part-
ners loss. Thus, the net benet canbe sharedby bothparties insome
equitable fashion (Goyal & Gupta, 1989).
When the demand is stochastic, lead time becomes an impor-
tant issue and its control leads to many benets. Shorter lead time
reduces the safety stock and the loss caused by stock-out, improves
customer service level and increases the competitive advantage of
business (Ouyang & Wu, 1997). As stated by Tersine (1994), lead
time usually comprises several components, such as order prepara-
tion, order transit, supplier lead time, delivery time and setup time.
In many practical situations, lead time can be reduced using an
0360-8352/$ - see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.cie.2009.04.018
* Corresponding author. Tel.: +91 512 2597101; fax: +91 512 2597553.
E-mail address: jkjha@iitk.ac.in (J.K. Jha).
Computers & Industrial Engineering 57 (2009) 10961104
Contents lists available at ScienceDirect
Computers & Industrial Engineering
j our nal homepage: www. el sevi er. com/ l ocat e/ cai e
added crashing cost. Inventory models incorporating lead time as a
decision variable were developed by several researchers. Liao and
Shyu (1991) rst devised a probabilistic inventory model in which
lead time was the unique decision variable. Later, several research-
ers (e.g., Ben-Daya & Raouf, 1994; Hariga & Ben-Daya, 1999; Lee,
Wu, & Hou, 2004; Moon & Choi, 1998; Ouyang & Chuang, 2001)
developed various analytical inventory models to explore the lead
time reduction problem. Ouyang, Chen and Chang (2002) extended
Moon and Chois (1998) model to include the possible relationship
between quality and lot size and then investigate the joint effects
of quality improvement and setup cost reduction in the model.
Lee (2005) extended the work of Ouyang and Chuang (2001) by
considering the mixtures of distribution of the lead time demand
and controllable backorder rate. Lee, Wu and Lei (2007) considered
the continuous review inventory systems with variable lead time
and backorder discount, where the ordering cost can be reduced
by capital investment. The underlying assumption in the above
studies was that lead time could be decomposed into several
mutually independent components, each with a different but xed
crash cost and so lead time crash cost is a function of the reduced
lead time. Also, other authors (e.g., Chang, Ouyang, Wu, & Ho,
2006; Ouyang & Chang, 2002; Pan & Hsiao, 2005; Pan, Hsiao, &
Lee, 2002; Pan, Lo, & Hsiao, 2004) explored the lead time reduction
problem assuming crash cost of lead time component as a function
of the order quantity and hence lead time crash cost is a function of
both the order quantity and the reduced lead time. The lead time
crash cost denition considered in the present study belongs to
the earlier category and lead time crash cost is the function of
the reduced lead time only.
The study of integrated inventory models in a single-vendor and
a single-buyer can be divided into two broad groups: the full cost
model and the service level approach model. In the full cost model,
the objective is to nd the optimal inventory policy which mini-
mizes the joint total relevant cost of the vendorbuyer integrated
system including the cost of shortages (Chang et al., 2006; Ouyang,
Wu, & Ho, 2004; Ouyang et al., 2007). The service level approach
introduces a service level constraint in place of the shortage cost
which implies that the stock-out level per cycle is bounded and
the availability of stock in a probabilistic or expected sense. There-
fore, many authors (Chu, Yang, & Chen, 2005; Lee, Wu, & Hsu, 2006;
Moon & Choi, 1994; Ouyang & Wu, 1997) replace the shortage cost
by a condition on the service level. All these authors focused on
determining the optimal policy of the buyer only under different
settings by optimizing the relevant cost of the buyer alone. How-
ever, in present study of the vendorbuyer integrated system, it is
worth investigating the optimal decisions of both the vendor and
the buyer by optimizing the cost of the vendor and the buyer to-
gether, when there is a service level constraint on the buyer.
Ouyanget al. (2004) extendedtheworkof PanandYang(2002) by
adding the shortage cost in the objective function and considering
reorder point as an additional decision variable. The present paper
modies the model of Ouyang et al. (2004) by introducing a service
level constraint, instead of considering the shortage cost in the
objective function. Specifying a service level avoids the difcult
practical issue of explicitly estimating the shortage cost. Moreover,
a service level criterion is generally easy to interpret and establish.
Therefore, SLC approach model is much more popular than full cost
model in spite of it being perceived as less mathematically tractable
(Chen & Krass, 2001).
In the previous works, the vendorbuyer integrated systems
with lead time crashing concept have been modeled as full cost
model including the cost of shortages. In this paper, we present
the SLC approach to the vendorbuyer integrated inventory model
involving variable lead time, i.e. the lead time is controllable and
reducible by adding additional crash costs. Consequently, this
study considers an integrated inventory model involving variable
lead time in which the objective is to minimize the joint total
expected cost of the vendorbuyer integrated system and subject
to a service level constraint on the buyer by simultaneously opti-
mizing the order quantity, lead time and the number of shipments
between the vendor and buyer in a production cycle. It is assumed
that the lead time demand follows a normal distribution. The fur-
ther assumptions which are made to describe the system and to
formulate the model are listed in the next section.
The remainder of this paper is organized as follows: Section 2
describes the notations and assumptions used throughout this
study. We formulate a vendorbuyer integrated inventory model
with service level issue and lead time reduction in Section 3. In
Section 4, a solution technique is developed to obtain the optimal
solution. A numerical example is provided in Section 5 to illustrate
the results. Finally, we draw some conclusions in Section 6.
2. Notations and assumptions
To establish the mathematical model, the following notations
and assumptions are used. Some additional notations will be listed
later when they are needed.
Notations
D average demand per unit time on the buyer
P production rate of the vendor, P > D
A buyers ordering cost per order
S vendors setup cost per setup
C
b
unit purchase cost paid by the buyer
C
v
unit production cost incurred by the vendor, C
v
< C
b
r
b
buyers holding cost rate (per monetary unit invested
in inventory) per unit time
r
v
vendors holding cost rate (per monetary unit
invested in inventory) per unit time
Q order quantity of the buyer (decision variable)
r reorder point of the buyer
L length of lead time for the buyer (decision variable)
a proportion of demands that are not met from stock
so (1 a) is the service level
B(r) expected demand shortage at the end of cycle
m the number of lots in which the product is delivered
from the vendor to the buyer in one production
cycle, a positive integer (decision variable)
X the lead time demand, which follows a normal
distribution with nite mean DL and standard
deviation r

L
p
; where r denotes the standard
deviation of demand per unit time, X NDL; r

L
p

E() mathematical expectation
x
+
maximum value of x and 0, i.e. x
+
= max {x, 0}
Assumptions
1. There is a single-vendor and a single-buyer and they deal with a
single item.
2. The buyer orders a lot of size Q and the vendor produces mQ
units with a nite production rate P (P > D) in one setup but
ships in quantity Q to the buyer over m times.
3. Buyer uses a continuous review inventory policy and the order
is placed whenever inventory level falls to the reorder point r.
4. Shortages are completely backordered.
5. The reorder point r = the expected demand during lead time
(DL) + safety stock (s), and s = k (standard deviation of lead
time demand), i.e. r DL kr

L
p
where k is the safety factor
and satises Pr(X > r) = q, q represents the allowable stock-out
probability during lead time.
J.K. Jha, K. Shanker / Computers & Industrial Engineering 57 (2009) 10961104 1097
6. The lead time L has n mutually independent components. The
ith component has a minimum duration a
i
, and normal duration
b
i
, and a crash cost per unit time c
i
. Further, we assume that
c
1
6 c
2
6 . . . 6 c
n
.
7. The lead time components are crashed one at a time starting
with the least c
i
component and so on.
8. Let L
0


n
j1
b
j
and L
i
be the length of lead time with compo-
nents 1, 2, . . ., i crashed to their minimum duration, then
L
i


n
ji1
b
j


i
j1
a
j
; i 1; 2; . . . ; n. The lead time crash cost
C(L) per cycle for a given L e [L
i
, L
i1
], is given by
CL c
i
L
i1
L

i1
j1
c
j
b
j
a
j
.
9. If a shortened lead time is requested then the extra costs
incurred by the vendor will be fully transferred to the buyer.
Therefore, lead time crash cost is the buyers cost component.
3. Model formulation
In this section a model is developed for the vendorbuyer inte-
grated system to minimize the joint total expected cost per unit
time of the vendorbuyer integrated system, which is the sum of
the ordering, holding and lead time crashing costs per unit time
for the buyer and the setup and holding costs per unit time for
the vendor, subject to a SLC on the buyer.
The buyer places an order after every Q demands, therefore for
average cycle time of Q/D, the expected ordering and lead time
crashing costs per unit time can be given by AD/Q and DC(L)/Q,
respectively. The expected net inventory level just before arrival
of a procurement is the safety stock, s = r DL. The expected net
inventory level immediately after arrival of a procurement is
Q + s. Hence, the average inventory over the cycle can be approxi-
mated by (Q/2) + s (Hadley & Whitin, 1963), i.e. Q=2 kr

L
p
(Assumption 5). So the buyers expected holding cost per unit time
is r
b
C
b
Q=2 kr

L
p
. Consequently, the total expected cost per unit
time for the buyer comprising of ordering cost, holding cost and
lead time crashing cost can be expressed as
TEC
b
Q; L
AD
Q
r
b
C
b
Q
2
kr

L
p
_ _

D
Q
CL: 1
The vendorbuyer integrated system is designed for a vendors
production situation in which once the buyer orders a lot of size Q,
the vendor begins production with a constant production rate P,
and a nite number of units are added to inventory until the pro-
duction run has been completed. The vendor produces the itemin a
lot of size mQ in each production cycle of length mQ/D, and the
buyer will receive the supply in m lots each of size Q. The rst
lot of size Q is ready for shipment after time Q/P just after the start
of the production, and then the vendor continues making the deliv-
ery on average every Q/D units of time until the inventory level
falls to zero (Fig. 1). The expected setup cost per unit time is given
by SD/(mQ).
Vendors average inventory is evaluated as the difference of the
vendors accumulated inventory and the buyers accumulated
inventory (see Fig. 1). That is,
mQ
Q
P
m1
Q
D
_ _

m
2
Q
2
2P
_ _

Q
2
D
12 m1
_ _ _ _
D
mQ

Q
2
m 1
D
P
_ _
1
2D
P
_ _
:
Therefore, the total expected cost per unit time for the vendor
comprising of setup cost and holding cost can be expressed as
TEC
v
Q; m
SD
mQ
r
v
C
v
Q
2
m 1
D
P
_ _
1
2D
P
_ _
: 2
Accordingly, the joint total expected cost per unit time for the
vendorbuyer integrated system is given by
JTECQ; L; m TEC
b
Q; L TEC
v
Q; m
JTECQ; L; m
D
Q
A
S
m
CL
_ _

Q
2
r
b
C
b
r
v
C
v
m 1
D
P
_ _
1
2D
P
_ _ _ _
r
b
C
b
kr

L
p
; L 2 L
i
; L
i1
: 3
3.1. Establishing the service level constraint
Usually it is not easy to quantify the penalty costs associated
with a shortage, as a stock-out event may include intangible inu-
ences. Therefore, several authors assume that buyer has set a target
service level corresponding to proportion of demands to be satis-
ed directly from available stock. Therefore, SLC puts a limit on
the proportion of demands not met from stock. For a given safety
factor which satises probability that lead time demand at the
buyer exceeds reorder point (Assumption 5), the actual proportion
of demands not met from stock should not exceed the desired va-
lue of a. Therefore, the SLC can be established as
Expecteddemand shortages at the endof cycle for a givensafety factor
Quantity available for satisfying the demandper cycle
6a; i:e:
Br
Q
6a:
Since the shortages occur when X > r, and then the expected de-
mand shortages at the end of cycle is given by, according to Ravin-
dran, Phillips and Solberg (1987), Br EX r

r

L
p
wk
where w(k) = /(k) k[1 U(k)] > 0, and /, Uare the standard nor-
mal probability density function and cumulative distribution func-
tion, respectively. Therefore, SLC is given by
r

L
p
wk
Q
6 a; L 2 L
i
; L
i1
: 4
Thus, the problem is to nd the optimal order quantity Q, lead
time L and the number of shipments in a production cycle m that
minimize the joint total expected cost (3), subject to SLC (4), that is
Min JTECQ; L; m
D
Q
A
S
m
CL
_ _

Q
2
r
b
C
b
r
v
C
v
m 1
D
P
_ _
1
2D
P
_ _ _ _
r
b
C
b
kr

L
p
Subject to
r

L
p
wk
Q
6 a
where L e [L
i
, L
i1
] and w(k) = /(k) k[1 U(k)] > 0.
4. Solution technique
The problem formulated in the previous section appears as con-
strained non-linear programming problem. To solve this problem,
we temporarily ignore the SLC and try to nd the optimal solution
of JTEC (Q, L, m). First, for xed m we take the rst partial deriva-
tives of JTEC (Q, L, m) with respect to Q and L e (L
i
, L
i1
), respec-
tively, and obtain
@JTECQ; L; m
@Q

D
Q
2
A
S
m
CL
_ _

1
2
r
b
C
b
r
v
C
v
m 1
D
P
_ _
1
2D
P
_ _ _ _
; 5
and
1098 J.K. Jha, K. Shanker / Computers & Industrial Engineering 57 (2009) 10961104
@JTECQ; L; m
@L

D
Q
c
i

r
b
2
C
b
krL
1=2
: 6
Hence, for xed m and L e [L
i
, L
i1
], JTEC (Q, L, m) is convex in Q,
since
@
2
JTECQ; L; m
@Q
2

2D
Q
3
A
S
m
CL
_ _
> 0:
However, for xed (Q, m), JTEC (Q, L, m) is concave in
L e [L
i
, L
i1
], because
@
2
JTECQ; L; m
@L
2

r
b
4
C
b
krL
3=2
< 0:
Therefore, for xed (Q, m), the minimum joint total expected
cost occurs at the end points of the interval [L
i
, L
i1
].
On the other hand, for xed m and L e [L
i
, L
i1
], we obtain opti-
mal order quantity Q = Q
JTEC
by setting (5) to zero as
Q
JTEC

2D A
S
m
CL
_ _
r
b
C
b
r
v
C
v
m 1
D
P
_ _
1
2D
P
_ _
_ _
1=2
; L 2 L
i
; L
i1
: 7
Thus, for xed m and L e [L
i
, L
i1
], when the SLC is ignored, (7)
gives optimal value of Q such that the joint total expected cost is
minimum. Now, the SLC (4) is taken into consideration and if
r=a

L
p
wk 6 Q for Q = Q
JTEC
, then Q
JTEC
is the local minimum of
JTEC (Q, L, m) and SLC is inactive. Otherwise, optimal value of Q
should be at least equal to Q
SLC
r=a

L
p
wk which is greater
than Q
JTEC
so that specied level of service can be achieved at min-
imum joint total expected cost and Q
SLC
is the local minimum of
JTEC (Q, L, m). Therefore, for xed m and L e [L
i
, L
i1
], the optimal
Q is given by max {Q
JTEC
, Q
SLC
}.
Next, in order to examine the effect of m on the joint total ex-
pected cost, we temporarily relax the integer requirement on m
and take the rst and second partial derivatives of JTEC (Q, L, m)
with respect to m and obtain
@JTECQ; L; m
@m

DS
Qm
2

Q
2
r
v
C
v
1
D
P
_ _
8
and
@
2
JTECQ; L; m
@m
2

2DS
Qm
3
> 0:
Therefore, JTEC (Q, L, m) is convex in m, for xed Q and
L e [L
i
, L
i1
]. As a result, the search for the optimal shipment num-
ber, m
*
, is reduced to nd a local minimum.
Now equating (8) to zero, we have
m
1
Q
2DS
r
v
C
v
1
D
P
_ _
_ _
1=2
: 9
From (9) it can be observed that Q and m have inverse relation-
ship but due to integer restriction on m this expression can not be
used for calculating the optimal value of m. Therefore, we suggest a
procedure in the following section to construct a range for search-
ing the value of optimal m.
Reorder point r
Q Q Q Q
Q Q
Q
B
u
y
e
r

s

s
t
o
c
k

Time
Safety stock s
L
Q
Buyers accumulated
inventory level
(m-1)Q/D
Q/D
Q
Q
V
e
n
d
o
r

s

s
t
o
c
k


Q
Q
Q
Q
Q
Q/P Q/D Q/D
Time
Production and shipment Shipment
Q
Production cycle
Q/D Q/D
Vendors accumulated
inventory level
mQ/D
mQ/P
mQ
Fig. 1. The inventory pattern for the vendor and the buyer.
J.K. Jha, K. Shanker / Computers & Industrial Engineering 57 (2009) 10961104 1099
4.1. Computation of range for nding the optimal number of shipments
During the process of crashing of the lead time components, the
service constraint may become active or inactive depending upon
the length of the lead time. Therefore, the following methods can
be used for producing the bounds on m, which reduces the compu-
tational effort considerably in searching the optimal number of
shipments in a production cycle of the vendor.
4.1.1. Case (a)
Service level constraint is inactive, i.e. r=a

L
p
wk 6 Q
JTEC
:
In this case optimal Q determined from (7) will satisfy the SLC.
For a particular value of L, substituting Q = Q
JTEC
from (7) into the
joint total expected cost expression (3), we get
JTECm 2D A
S
m
CL
_ _
r
b
C
b
r
v
C
v
m 1
D
P
_ _
1
2D
P
_ _ _ _ _ _
1=2
r
b
C
b
kr

L
p
:
10
Ignoring the terms that are independent of m, and taking the
square of (10), then minimizing JTEC (m) is equivalent to
minimizing
JTECm
2
2D A
S
m
CL
_ _
r
b
C
b
r
v
C
v
m 1
D
P
_ _
1
2D
P
_ _ _ _
2D ACL r
b
C
b
r
v
C
v
1
2D
P
_ _ _ _
Sr
v
C
v
1
D
P
_ _ _
mr
v
C
v
ACL 1
D
P
_ _

S
m
r
b
C
b
r
v
C
v
1
2D
P
_ _ _ __
:
Once again, ignoring the terms that are independent of m, the
minimization of the problem can be reduced to that of minimizing
Zm mr
v
C
v
A CL 1
D
P
_ _

S
m
r
b
C
b
r
v
C
v
1
2D
P
_ _ _ _
: 11
The optimal value of m = m
JTEC
is obtained when
Zm
JTEC
6 Zm
JTEC
1 and Zm
JTEC
6 Zm
JTEC
1: 12
From (11) and (12), we get
m
JTEC
m
JTEC
1 6
S r
b
C
b
r
v
C
v
1
2D
P
_ _ _
r
v
C
v
A CL 1
D
P
_ _
and
m
JTEC
m
JTEC
1 P
S r
b
C
b
r
v
C
v
1
2D
P
_ _ _
r
v
C
v
A CL 1
D
P
_ _ :
Therefore, m
JTEC
satises
m
JTEC
m
JTEC
1 6
S r
b
C
b
r
v
C
v
1
2D
P
_ _ _
r
v
C
v
A CL 1
D
P
_ _ 6 m
JTEC
m
JTEC
1: 13
4.1.2. Case (b)
Service level constraint is active, i.e. r=a

L
p
wk > Q
JTEC
:
When the SLC is active, the optimal order quantity should be at
least enough to satisfy the service level requirement. Therefore, by
putting Q
SLC
r=a

L
p
wk for Q in joint total expected cost
expression (3) and adopting the procedure similar to that detailed
in case (a), the optimal value of m = m
SLC
will satisfy
m
SLC
m
SLC
1 6
2DSa
2
r
2
Lr
v
C
v
1
D
P
_ _
wk
2
6 m
SLC
m
SLC
1: 14
Based on the degree of crashing of the lead time the bounds on
m can be calculated using (13) and (14). Let the bounds on m
from (13) and (14) be given by m
lower
JTEC
; m
upper
JTEC
and m
lower
SLC
; m
upper
SLC
,
respectively. Then, if none of the lead time components crashes,
the lead time length will be maximum and the corresponding
crash cost is zero, which give m
lower
SLC
and m
upper
JTEC
. On the other hand,
if all the components of the lead time are crashed to their mini-
mum limits, the lead time length will be minimum and the cor-
responding crash cost is maximum, which give m
upper
SLC
and m
lower
JTEC
.
Since the optimal order quantity is given by max {Q
JTEC
, Q
SLC
}
and (9) shows that m and Q have inverse relationship. Therefore,
during the entire process of crashing of the lead time components
to their minimum limits the optimal m will be located purely in
m
lower
SLC
; m
upper
SLC
_
if SLC remains active throughout or in
m
lower
JTEC
; m
upper
JTEC
_ _
if SLC remains inactive throughout, and these
two ranges will be non-overlapping. On the other hand, it is quite
possible that SLC may remain both active and inactive during the
process of crashing, then it is difcult to say in which of the
ranges the optimal m will fall and the two ranges will be overlap-
ping, i.e. at least one element will be in common. In this situation
the optimal m will fall in a range resulting from the union of
these two ranges which contains the elements of both the ranges,
i.e. m
lower
SL
; m
upper
SL
_
[ m
lower
TC
; m
upper
TC
_
. Thus, the rule for nding the
range for searching the optimal shipment number
m
*
e [m
lower
, m
upper
] can be summarized as
If m
upper
JTEC
< m
lower
SLC
;
m
lower
m
lower
JTEC
m
upper
m
upper
JTEC
_
If m
upper
SLC
< m
lower
JTEC
;
m
lower
m
lower
SLC
m
upper
m
upper
SLC
_
Otherwise ;
m
lower
minfm
lower
JTEC
; m
lower
SLC
g
m
upper
maxfm
upper
JTEC
; m
upper
SLC
g
_
With the help of the method described above the optimal value
of Q can be determined for xed m and L e [L
i
, L
i1
]. Further, based
on the convexity and concavity behavior of objective function with
respect to the decision variables the following algorithm is devel-
oped to nd the optimal values for the order quantity, lead time
and the number of deliveries in one production cycle.
Algorithm
Step 1: Compute the range of m as per the rule discussed above.
Step 2: Set m = m
lower
.
Step 3: For each L
i
, i = 0, 1, . . ., n, perform (3.1) to (3.3).
(3.1) Compute Q
i
JTEC
from Eq. (7).
(3.2) Let Q
i
maxfQ
i
JTEC
; r=a

L
i
p
wkg.
(3.3) Compute the corresponding JTEC (Q
i
, L
i
, m) by
putting Q = Q
i
in Eq. (3).
Step 4: Find min
i=0,1,. . .,n
JTEC (Q
i
, L
i
, m). Let JTECQ

m
; L

m
; m
min
i0;1;...;n
JTECQ
i
; L
i
; m, then Q

m
; L

m
is the optimal solu-
tion for xed m.
Step 5: Set m = m + 1, repeat Steps 3 and 4 to get
JTEC Q

m
; L

m
; m
_ _
.
Step 6: If JTEC Q

m
; L

m
; m
_ _
6 JTEC Q

m1
; L

m1
; m 1
_ _
, then
go to Step 5, otherwise go to Step 7.
Step 7: Set Q

; L

; m

m1
; L

m1
; m 1
_ _
, then (Q
*
, L
*
, m
*
)
is the optimal solution.
5. Numerical example
Consider a two echelon supply chain inventory system with the
following data: D = 1000 units/year, P = 3200 units/year, A = $25/
1100 J.K. Jha, K. Shanker / Computers & Industrial Engineering 57 (2009) 10961104
order, S = $400/setup, C
b
= $25/unit, C
v
= $20 per unit, r = 7 units/
week, the service level, 1 a = 0.985, r
b
= 0.2, r
v
= 0.2 and the lead
time has three components with data shown in Table 1.
Applying the procedure of the proposed algorithm for different
k = 0.00, 0.25, 0.50, 0.75, 1.00, 1.25 and 1.50, the resulting bounds
on m and solution procedure are illustrated in Tables 2 and 3,
respectively. Since the expression in (13) is independent of k and
so we always get the same bounds on m 2 m
lower
JTEC
; m
upper
JTEC
_ _
for all
values of k. For k = 0.75, the optimal solution from Table 3 can be
read off as Q = 149.87 units, L = 6 weeks, m = 4 and the correspond-
ing joint total expected annual cost as $1994.25. In Table 3, it can
be noted that for k = 1.00, the convexity behavior of total cost
expression with respect to m stops to check the value of JTEC at
higher values of m in the range, which is mentioned in Step 6 of
the algorithm.
5.1. Effects of service level on JTEC
In order to see the effects of service level, we summarize the
computational results for different a by varying safety factor in
Table 4. When the order quantity determined from joint total
expected cost expression (3) is not enough to satisfy the desired
service level for a chosen value of safety factor, then the SLC be-
comes active and the optimal order quantity takes the value from
SLC (4). Fig. 2 depicts that for smaller safety factor, JTEC decreases
Table 1
Lead time data.
Lead time component i Normal duration
b
i
(days)
Minimum duration
a
i
(days)
Unit crashing
cost c
i
($/day)
1 20 6 0.1
2 20 6 1.2
3 16 9 5.0
Table 2
Results for bounds on m.
k m
lower
JTEC
m
upper
JTEC
m
lower
SLC
m
upper
SLC
m
lower
m
upper
0.00 3 5 1 2 1 2
0.25 3 5 2 2 2 2
0.50 3 5 2 3 2 5
0.75 3 5 3 5 3 5
1.00 3 5 5 8 3 8
1.25 3 5 8 13 3 5
1.50 3 5 14 23 3 5
Table 3
Illustration of the solution procedure for different safety factor k (lead time in weeks).
m i = 0, L
i
= 8 i = 1, L
i
= 6 i = 2, L
i
= 4 i = 3, L
i
= 3
Q, JTEC (Q, L, m) Q, JTEC (Q, L, m) Q, JTEC (Q, L, m) Q, JTEC (Q, L, m)
k = 0.00 1 526.47, 2452.48 455.94, 2360.02 376.60, 2353.72 391.18, 2444.89
m
lower
= 1 2 526.47, 2796.49 455.94, 2548.28 372.27, 2328.51 322.40, 2313.70
k = 0.25 2 377.85, 2320.56 327.23, 2185.84 267.18, 2130.06 248.64, 2252.92
m
lower
= 2
k = 0.50 2 260.98, 2086.05 226.02, 2061.64 232.47, 2127.27 248.64, 2268.08
m
lower
= 2 3 260.98, 2189.46 226.02, 2077.45 184.54, 2075.79 189.75, 2259.89
k = 0.75 3 173.05, 2005.87 164.89, 2001.75 173.34, 2089.29 189.75, 2275.05
m
lower
= 3 4 173.05, 2051.20 149.87, 1994.25 140.54, 2090.34 156.78, 2318.75
5 173.05, 2173.57 149.87, 2066.86 122.37, 2114.72 135.43, 2381.68
k = 1.00 3 164.17, 2027.94 164.89, 2023.19 173.34, 2106.79 189.75, 2290.20
m
lower
= 3 4 131.31, 2002.94 132.04, 2000.31 140.54, 2107.84 156.78, 2333.90
5 110.34, 2002.28 111.07, 2001.67 119.52, 2131.65 135.43, 2396.84
k = 1.25 3 164.17, 2052.69 164.89, 2044.62 173.34, 2124.29 189.75, 2305.36
m
lower
= 3 4 131.31, 2027.69 132.04, 2021.74 140.54, 2125.34 156.78, 2349.06
5 110.34, 2027.03 111.07, 2023.10 119.52, 2149.15 135.43, 2411.99
k = 1.50 3 164.17, 2077.44 164.89, 2066.05 173.34, 2141.79 189.75, 2320.51
m
lower
= 3 4 131.31, 2052.44 132.04, 2043.17 140.54, 2142.84 156.78, 2364.21
5 110.34, 2051.78 111.07, 2044.53 119.52, 2166.65 135.43, 2427.15
Table 4
Computational results for different values of a (lead time in weeks).
a k L
*
m
*
Q
*
JTEC ()
0.055 0.00 8 4 143.58 1911.55
0.25 8 5 110.34 1928.04
0.50 8 5 110.34 1952.78
0.75 8 5 110.34 1977.53
1.00 6 4 132.04 2000.31
1.25 6 4 132.04 2021.74
1.50 6 4 132.04 2043.17
0.035 0.00 6 3 195.4 1965.45
0.25 6 4 140.24 1939.49
0.50 8 5 111.85 1952.96
0.75 8 5 110.34 1977.53
1.00 6 4 132.04 2000.31
1.25 6 4 132.04 2021.74
1.50 6 4 132.04 2043.17
0.025 0.00 6 2 273.56 2058.63
0.25 6 3 196.34 1988.48
0.50 6 4 135.61 1958.12
0.75 8 5 110.34 1977.53
1.00 6 4 132.04 2000.31
1.25 6 4 132.04 2021.74
1.50 6 4 132.04 2043.17
0.015 0.00 3 2 322.40 2313.70
0.25 4 2 267.18 2130.06
0.50 6 2 226.02 2061.64
0.75 6 4 149.87 1994.25
1.00 6 4 132.04 2000.31
1.25 6 4 132.04 2021.74
1.50 6 4 132.04 2043.17
0.010 0.00 3 1 483.59 2500.08
0.25 4 1 400.77 2375.78
0.50 4 2 276.82 2159.23
0.75 6 2 224.80 2083.02
1.00 6 4 142.77 2006.16
1.25 6 4 132.04 2021.74
1.50 6 4 132.04 2043.17
0.005 0.00 3 1 967.19 3516.89
0.25 3 1 694.16 2873.30
0.50 3 1 479.37 2525.90
0.75 3 2 317.92 2351.16
1.00 4 2 233.11 2162.28
1.25 6 3 173.35 2047.05
1.50 6 4 132.04 2043.17
J.K. Jha, K. Shanker / Computers & Industrial Engineering 57 (2009) 10961104 1101
nonlinearly and then it starts increasing linearly, because beyond
certain value of the safety factor SLC becomes inactive and the
optimal ordering quantity determined from joint total expected
cost expression is more than enough to satisfy the desired service
requirement. Now, as long as SLC remains active with increase in
safety factor for a given service level, the expected shortages per
cycle goes on decreasing which requires the order quantity to de-
crease. Since JTEC is convex in Q and so any decrease in Q to the
right side of Q
JTEC
causes nonlinear decrease in JTEC. With further
increase in safety factor the requirement of Q should decrease for
a desired service level, but it is not allowed to drop below Q
JTEC
and so SLC becomes inactive and the order quantity becomes con-
stant at Q
JTEC
. Therefore, increasing the safety factor for inactive SLC
only increases holding cost of the safety stock linearly and hence
the JTEC. It can be observed that for any service level, once the joint
total expected annual cost starts varying linearly with safety factor,
the joint total expected annual cost becomes independent of ser-
vice level and takes the same value for larger values of safety fac-
tor. Also, at k = 0.25 raising the service level by 2.0% from 94.5 to
96.5 requires an additional cost of $11.45, whereas raising the ser-
vice level by 0.5% from 99.0 to 99.5 requires an additional cost of
$497.52. Thus, the required JTEC grows rapidly with an increase
in the desired service level.
5.2. Effects of holding costs of vendor and buyer on JTEC
Theabovenumerical examplehas beensolvedbytakingthesame
holding cost rate per unit time for buyer and vendor both. However,
inpractice thevendors holdingcost rate per unit time, r
v
, maynot be
equal to the buyers holding cost rate per unit time, r
b
. Therefore, the
effect of holding cost of vendor and buyer on the joint total expected
cost needs to be examined. The results are summarizedinTable 5for
the different values of r
b
and r
v
. When r
v
< r
b
, Table 5 shows that the
buyer tends to order small quantity by maintaining the relatively
higher safety stock to meet the desired service level and the vendor
tends to produce more quantity in one production setup. This result
reveals that if the buyer has a higher holding cost rate, then he tends
tomaintainhigher safety stockandorder smaller quantityeachtime
for keeping his average inventory level as low as possible. On the
other hand, if the vendor has a lower holding cost rate, he will pro-
duce in bigger lot size in one production setup. On the converse,
when r
v
> r
b
, Table 5 shows that the vendor prefers to produce in
smaller lot size in each production setup so that he could be able
to keep his average inventory level as low as possible.
6. Independent inventory policy of buyer and vendor
This sectiondiscusses about inventorypolicyinwhichbothbuyer
and vendor minimize their own cost independently. The buyer com-
putes his economic order quantity using (1) and SLC (4). To obtain
the minimum cost lot size, take the partial derivatives of TEC
b
(Q, L)
with respect to Q and L in each time interval (L
i
, L
i1
), thus
@TEC
b
Q; L
@Q

D
Q
2
A CL
r
b
2
C
b
; 15
and
@TEC
b
Q; L
@L

D
Q
c
i

r
b
2
C
b
krL
1=2
: 16
Hence, for xed L e [L
i
, L
i1
], TEC
b
(Q, L) is convex in Q, since
@
2
TEC
b
Q; L
@Q
2

2D
Q
3
A CL > 0:
However, for xed Q, TEC
b
(Q, L) is concave in L e [L
i
, L
i1
], because
@
2
TEC
b
Q; L
@L
2

r
b
4
C
b
krL
3=2
< 0:
Therefore, for xed Q, the minimum total expected cost per unit
time for the buyer will occur at the end points of the interval
[L
i
, L
i1
]. Now setting (15) to zero, we have
Q
2DA CL
r
b
C
b
_ _
1=2
; L 2 L
i
; L
i1
: 17
Thus, for xed L e [L
i
, L
i1
], when the service level constraint is ig-
nored, (17) gives optimal value of Q such that the total expected
cost per unit time for the buyer has a minimum value. Now consid-
ering the service level constraint, the optimal Q can be obtained as
Q max
2DA CL
r
b
C
b
_ _
1=2
;
r
a

L
p
wk
_ _
; L 2 L
i
; L
i1
: 18
Now, for all possible values of lead time and its corresponding
crashing cost, the order quantity and the corresponding total ex-
pected cost of the buyer are obtained using (18) and (1), respec-
tively. Then, the order quantity, lead time will have the optimal
value obtained corresponding to the total expected cost of the buyer
for which it has minimum value. On the other hand, total expected
cost of the vendor which is given by (2) that can be minimized by
Table 5
Computational results for different values of r
b
and r
v
(lead time in weeks).
r
b
r
v
k
*
L
*
m
*
Q
*
JTEC ()
0.10 0.10 0.61 6 4 189.62 1380.12
0.15 0.55 6 3 208.89 1578.12
0.20 0.39 6 2 267.27 1732.43
0.15 0.10 0.75 6 5 149.87 1498.16
0.15 0.73 6 4 155.12 1705.26
0.20 0.65 6 3 177.53 1873.58
0.20 0.10 0.85 6 6 125.64 1598.86
0.15 0.85 6 5 125.64 1810.99
0.20 0.81 6 4 134.94 1984.47
1800
2100
2400
2700
3000
3300
3600
0.00 0.25 0.50 0.75 1.00 1.25 1.50
Safety factor ( k)
J
T
E
C

(
$
)
Service Level=94.5%
Service Level=96.5%
Service Level=97.5%
Service Level=98.5%
Service Level=99.0%
Service Level=99.5%
Fig. 2. Variation of joint total expected cost with safety factor for different service
level.
1102 J.K. Jha, K. Shanker / Computers & Industrial Engineering 57 (2009) 10961104
taking the optimal order quantity and lead time already determined
for the buyer and selecting suitable integer value of m.
We summarize the computational results of independent opti-
mization and joint optimization in Table 6 for different a. As the
Fig. 2 shows that for any service level, there exists a unique value
of the safety factor at which the system has lowest total cost.
Therefore, results are compared in both the approaches only for
the best value of the safety factor corresponding to a given service
level. Table 6 shows that for the same service level in both the ap-
proaches, the independent optimization model requires the buyer
to order in smaller lot size and to keep higher safety factor, and
simultaneously the vendor to produce in a bigger lot size. In inde-
pendent optimization approach, the buyer has privilege to decide
the optimal order lot size and so he takes advantage of reducing
the holding cost by lowering his average cycle inventory level,
but achieve the desired service level by maintaining relatively
higher safety factor. Now the vendor has to decide his optimal pro-
duction lot size only in the integer multiple of the optimal order lot
size favored by the buyer, which increases vendors holding cost.
Table 6 shows that the total cost of the vendor increases in inde-
pendent optimization. Therefore, vendors loss in holding cost is
more than the gain in his setup cost. Further, it can be observed
that the joint optimization approach gives lower total cost of the
vendorbuyer integrated system. Therefore it can be concluded
that the gain of buyer is small as compared to the loss of vendor
in independent optimization strategy and joint optimization model
results in net savings to the integrated system.
7. Conclusions
In this study, we consider two echelon supply chain problem
consisting of a single-vendor and a single-buyer. Previous works
on lead time reduction for integrated vendorbuyer problem have
been modeled by taking the shortage cost. Here, the same system
has been studied taking service level as a constraint in place of
the shortage cost. An efcient procedure has been suggested to nd
the bounds on shipment number for searching the optimal ship-
ment number, which reduces the considerable computational ef-
fort. Further, we developed an algorithm to minimize the joint
total expected cost of the vendorbuyer integrated system by
simultaneously optimizing the ordering quantity, lead time and
the number of shipments fromthe vendor to the buyer in a produc-
tion cycle. It has been shown that if service level is a binding con-
straint, then the joint total expected cost decreases nonlinearly
with safety factor otherwise increases linearly. Finally, the results
of the numerical example indicate that by adopting the joint opti-
mization model, the gain of vendor exceeds the loss of buyer,
which gives the net savings to the integrated system.
Acknowledgment
The authors greatly appreciate the anonymous referees for their
valuable and helpful suggestions to improve an earlier version of
the paper.
References
Banerjee, A. (1986). A joint economic-lot-size model for purchaser and vendor.
Decision Sciences, 17(3), 292311.
Ben-Daya, M., & Raouf, A. (1994). Inventory models involving lead time as a decision
variable. Journal of the Operational Research Society, 45(5), 579582.
Chang, H. C., Ouyang, L. Y., Wu, K. S., & Ho, C. H. (2006). Integrated vendorbuyer
cooperative inventory models with controllable lead time and ordering cost
reduction. European Journal of Operational Research, 170(2), 481495.
Chen, F. Y., & Krass, D. (2001). Inventory models with minimal service level
constraints. European Journal of Operational Research, 134(1), 120140.
Chu, P., Yang, K. L., & Chen, P. S. (2005). Improved inventory models with service
level and lead time. Computers and Operations Research, 32(2), 285296.
Goyal, S. K. (1976). An integrated inventory model for a single suppliersingle
customer problem. International Journal of Production Research, 15(1), 107111.
Goyal, S. K. (1988). A joint economic-lot-size model for purchaser and vendor: A
comment. Decision Sciences, 19(1), 236241.
Goyal, S. K. (2003). A note on: On controlling the controllable lead time component
in the integrated inventory models. International Journal of Production Research,
41(12), 28732875.
Goyal, S. K., & Gupta, Y. P. (1989). Integrated inventory models: The buyervendor
coordination. European Journal of Operational Research, 41(3), 261269.
Hadley, G., & Whitin, T. (1963). Analysis of inventory systems. Englewood Cliffs, New
Jersey: Prentice Hall.
Hariga, M., & Ben-Daya, M. (1999). Some stochastic inventory models with
deterministic variable lead time. European Journal of Operational Research,
113(1), 4251.
Hoque, M. A., & Goyal, S. K. (2006). A heuristic solution procedure for an integrated
inventory system under controllable lead-time with equal or unequal sized
batch shipments between a vendor and a buyer. International Journal of
Production Economics, 102(2), 217225.
Lee, W. C. (2005). Inventory model involving controllable backorder rate and
variable lead time demand with the mixtures of distribution. Applied
Mathematics and Computation, 160(3), 701717.
Lee, W. C., Wu, J. W., & Hou, W. B. (2004). A note on inventory model involving
variable lead time with defective units for mixtures of distribution. International
Journal of Production Economics, 89(1), 3144.
Lee, W. C., Wu, J. W., & Hsu, J. W. (2006). Computational algorithm for inventory
model with a service level constraint, lead time demand with the mixture of
distributions and controllable negative exponential backorder rate. Applied
Mathematics and Computation, 175(2), 11251138.
Lee, W. C., Wu, J. W., & Lei, C. L. (2007). Computational algorithmic procedure for
optimal inventory policy involving ordering cost reduction and back-order
discounts when lead time demand is controllable. Applied Mathematics and
Computation, 189(1), 186200.
Liao, C. J., & Shyu, C. H. (1991). An analytical determination of lead time with normal
demand. International Journal of Operations and Production Management, 11(9),
7278.
Moon, I., & Choi, S. (1994). The distribution free continuous review inventory
system with a service level constraint. Computers and Industrial Engineering,
27(14), 209212.
Moon, I., & Choi, S. (1998). A note on lead time and distributional assumptions in
continuous review inventory models. Computers and Operations Research,
25(11), 10071012.
Ouyang, L. Y., & Chang, H. C. (2002). Lot size reorder point inventory model with
controllable lead time and set-up cost. International Journal of Systems Science,
33(8), 635642.
Ouyang, L. Y., Chen, C. K., & Chang, H. C. (2002). Quality improvement, setup
cost and lead-time reductions in lot size reorder point models with an
imperfect production process. Computers and Operations Research, 29(12),
17011717.
Ouyang, L. Y., & Chuang, B. R. (2001). Mixture inventory model involving variable
lead time and controllable backorder rate. Computers and Industrial Engineering,
40(4), 339348.
Ouyang, L. Y., & Wu, K. S. (1997). Mixture inventory model involving variable lead
time with a service level constraint. Computers and Operations Research, 24(9),
875882.
Ouyang, L. Y., Wu, K. S., & Ho, C. H. (2004). Integrated vendorbuyer cooperative
models with stochastic demand in controllable lead time. International Journal
of Production Economics, 92(3), 255266.
Table 6
Summary of independent optimization vs. joint optimization (lead time in weeks).
a Independent optimization Joint optimization Savings
k
*
L
*
m
*
Q
*
TEC
b
TEC
v
JTEC
I
k
*
L
*
m
*
Q
*
JTEC TEC
b
TEC
v
0.005 1.46 6 5 109.93 640.15 1401.06 2041.21 1.37 6 4 134.18 2032.28 649.65 1382.63 8.93
0.010 1.13 6 5 110.77 612.14 1400.68 2012.82 1.03 6 4 134.80 2003.29 621.15 1382.14 9.53
0.015 0.91 6 5 112.65 594.00 1400.14 1994.14 0.81 6 4 134.94 1984.47 602.44 1382.03 9.67
0.020 0.75 6 5 112.40 580.17 1400.19 1980.36 0.64 6 4 135.37 1970.04 588.31 1381.73 10.32
0.025 0.61 6 5 113.77 568.77 1400.01 1968.78 0.50 6 4 135.61 1958.12 576.57 1381.55 10.66
0.030 0.49 6 5 114.78 558.97 1400.01 1958.98 0.38 6 4 135.65 1947.85 566.32 1381.53 11.13
Note: JTEC
I
denotes the total expected cost of system in independent optimization approach, which is equal to TEC
b
+ TEC
v
.
J.K. Jha, K. Shanker / Computers & Industrial Engineering 57 (2009) 10961104 1103
Ouyang, L. Y., Wu, K. S., & Ho, C. H. (2007). An integrated vendorbuyer inventory
model with quality improvement and lead time reduction. International Journal
of Production Economics, 108(12), 349358.
Pan, J. C. H., & Hsiao, Y. C. (2005). Integrated inventory models with controllable
lead time and backorder discount considerations. International Journal of
Production Economics, 9394, 387397.
Pan, J. C. H., Hsiao, Y. C., & Lee, C. J. (2002). Inventory models with xed and variable
lead time crash costs considerations. Journal of the Operational Research Society,
53(9), 10481053.
Pan, J. C. H., Lo, M. C., & Hsiao, Y. C. (2004). Optimal reorder point inventory models
with variable lead time and backorder discount considerations. European
Journal of Operational Research, 158(2), 488505.
Pan, J. C. H., & Yang, J. S. (2002). A study of an integrated inventory with controllable
lead time. International Journal of Production Research, 40(5), 12631273.
Ravindran, A., Phillips, D. T., & Solberg, J. J. (1987). Operations research: Principles and
practice. New York: Wiley.
Tersine, R. J. (1994). Principles of inventory and materials management. Englewood
Cliffs, New Jersey: Prentice Hall.
Yang, J. S., & Pan, J. C. H. (2004). Just-in-time purchasing: An integrated inventory
model involving deterministic variable lead time and quality improvement
investment. International Journal of Production Research, 42(5), 853863.
1104 J.K. Jha, K. Shanker / Computers & Industrial Engineering 57 (2009) 10961104

You might also like