Optimal inventory model with defective units involving controllable backorder

rate and variable lead time with mixtures of distribution

Wen-Chuan Lee
1
, Jong-Wuu Wu
2*
and Hsin-Hui Tsou
2
1
Department of International Business, Chang Jung Christian University, Tainan 71101, Taiwan, R.O.C.
2
Department of Applied Mathematics, National Chiayi University, Chiayi City 60004, Taiwan, R.O.C.
Abstract
This paper considers that the number of defec-
tive units in an arrival order to be a binominal
random variable. We derive a modified mixture
inventory model with backorders and lost sales, in
which the order quantity and lead time are deci-
sion variables. In our studies, we also assume that
the backorder rate is dependent on the length of
lead time through the amount of shortages and let
the backorder rate is a control variable. In addition,
we assume that the lead time demand follows a
mixtures of normal distribution, and then relax the
assumption about the form of the mixtures of dis-
tribution function of the lead time demand and
apply the minimax distribution free procedure to
solve the problem. Furthermore, we develop an
algorithm procedure to obtain the optimal ordering
strategy for each case. Finally, two numerical ex-
amples are also given to illustrate the results.
Keywords: Defective units; Order quantity;
Backorder rate; Minimax distribution free proce-
dure; Mixtures of distribution
1. Introduction
In most of the literatures dealing with inventory
problems, either in deterministic or probabilistic
model, lead time is viewed as a prescribed con-
stant or a stochastic variable, which, therefore, is
not subject to control (see, e.g., Naddor, 1966;
Silver and Peterson, 1985). In fact, lead time usu-
ally consists of the following components: order
preparation, order transit, supplier lead time, de-
livery time, and setup time (Tersine, 1982). In
many practical situations, lead time can be re-
duced by an added crashing cost; in other words, it
is controllable. By shortening the lead time, we
can lower the safety stock (SS), reduce the loss
*
Correspondence: Jong-Wuu Wu, Department of
Applied Mathematics, National Chiayi University,
300 Syuefu RD., Chiayi City 60004, Taiwan,
R.O.C.; E-mail: jwwu@mail.ncyu.edu.tw.
caused by stock-out, improve the service level to
the customer, and increase the competitive ability
in business.
Liao and Shyu (1991) first presented a prob-
abilistic model in which the order quantity is pre-
determined and lead time is a unique decision
variable. Later, Ben-Daya and Raouf (1994) ex-
tended the Liao and Shyu's model by considering
both the lead time and the order quantity as deci-
sion variables. In their studies (Liao and Shyu,
1991; Ben-Daya and Raouf, 1994), they assume
that the probability of allowable stock-out during
the lead time is very small, and hence the short-
ages are neglected. In some recent papers, Ouyang
et al. (1996), Ouyang and Chuang (2001) and Wu
and Tsai (2001) have also extended Ben-Daya and
Raoufs model by considering the stock-out com-
ponent.
The above body of literature does not describe
the possible relationship between the order lot and
quantity. As a result of imperfect production of the
supplier, and/or damage in transit, it is often that
an arrival order lot may contain some defective
units. The presence of defective units in orders
would have an impact on the number of shortages
and the frequency of orders in the inventory sys-
tem. Therefore, ordering policy determined by
conventional inventory models may not be appro-
priate when applied to the defective inventory
situations.
Hence, Ouyang and Wu (1999) consider both
lead time and the order quantity as the decision
variables of a mixture of backorders and lost sales
inventory model in which the shortages are al-
lowed and consider the demand of the lead time
with normal distribution. Inventory is continu-
ously reviewed, an order of size Q is made when-
ever the inventory level (based on the number of
non-defective items) falls to the reorder point r.
Besides, they assume that an arrival order may
contain some defective units, and the number of
defective units in an arrival order of size Q is a
random variable with binomial probability distri-
bution. In addition, we also assume that the pur-
chaser inspects the entire items and the defective
units in each lot will be returned to the vendor at
T h e 2 6 t h W o r k s h o p o n C o m b i n a t o r i a l M a t h e m a t i c s a n d C o m p u t a t i o n T h e o r y
3 1 4
the time of delivery of the next lot. Therefore, the
model will have an extra cost in the inspection of
each lot and an extra cost for holding the defective
units in stock until the time they are returned to
the supplier.
In this study, we seek to extend the model of
Ouyang and Wu (1999) with the fixed rate propose
a more general model that allows the backorder
rate as a control variable. According to the opinion
of Ouyang and Chuang (2001) (also see Lee
(2005)), under most market behaviors, they think
that many products of famous brands or fashion-
able commodities may lead to a situation in which
customers prefer their demands to be backordered
while shortages occur. Certainly, if the quantity of
shortages is accumulated to a degree that exceeds
the waiting patience of customers, some may re-
fuse the backorder case. This phenomenon reveals
that, as shortages occur, the longer the length of
lead time is, the larger the amount of shortages is,
the smaller the proportion of customers can wait,
and hence the smaller the backorder rate would be.
Hence, they assumed that the backorder rate is
dependent on the length of lead time through the
amount of shortages. So, this paper consider a
continuous review inventory model with the
backorder rate that proposed by using the idea of
Ouyang and Chuang (2001) (also see Lee (2005))
to present a new general form.
However, when the demands of the different
customers are not identical in the lead time, then
we cannot use only a single distribution (such as
Ouyang and Wu, 1999; Ouyang et al., 1999) to
describe the demand of the lead time. Hence, we
consider the mixtures of distribution (see Everitt
and Hand, 1981). We first assume that the lead
time demand follows a mixture of normal distribu-
tion and develop an algorithm procedure to find
the optimal order quantity, the optimal lead time
and the optimal reorder point. And then we con-
sider any mixture of distribution function (d.f.),
say F
*
= pF
1
+ (1 p)F
2
, of the lead time demand
has only known finite first and second moments
(and hence, mean and variance are also known and
finite) but make no assumption on the distribution
form of F
*
. That is, F
1
and F
2
of F
*
belong to the
class O of all single d.f.s with finite mean and
variance. Our goal is to solve a mixture inventory
model with defective units by using the minimax
distribution free approach. The minimax distribu-
tion free approach for our inventory model is to
find the most unfavorable d.f.s F
1
and F
2
of F
*
for
each decision variable and then to minimize over
the decision variables. Furthermore, the purpose of
this paper is to develop an algorithm procedure for
the mixture inventory model with defective units
to find the optimal order quantity, optimal lead
time and the optimal reorder point when the dis-
tribution of the lead time demand is mixtures of
normal distribution or mixtures of free distribution.
Finally, two numerical examples are also given to
illustrate that when 0 p = or 1, the model con-
siders only one kind of customers demand; when
0 1 p < < , the model considers two kinds of cus-
tomers demand for the fixed fraction of the de-
mand backordered | and defective rate u . It
implies that the minimum expected total annual
cost of two kinds of customers demand are larger
than the minimum expected total annual cost of
one kind of customers demand. Thus, the mini-
mum expected total annual cost increases as the
distance between p and 0 (or 1) increases for the
fixed | and u . Hence, if the true distribution of
the lead time demand is mixtures of normal dis-
tribution or mixtures of free distribution, we use a
single distribution (such as Ouyang and Wu, 1999;
Ouyang et al., 1999) to substitute the true distribu-
tion of the lead time demand; then the minimum
expected total annual cost will be underestimated.
2. Notations and assumptions
To establish the mathematical model, the nota-
tions and non-negative cost parameters are exactly
the same as those in Ouyang and Wu (1999) ex-
cept the following notations:
X the lead time demand which has a mixture of
d.f. F
*
= pF
1
+ (1 p)F
2
with finite mean
*
L and standard deviation
*
( 0 ) L o > ,
a random variable.
x
+
maximum value of x and 0, i.e.,
{ } max , 0 x x
+
= .
x

maximum value of x and 0,

i.e., { } max , 0 x x

= .
r reorder point.
( 0 )
1 , 0 ,
0 , o.w.
X r
x r
I
< <
< <
=

The assumptions of the model are exactly the

same as those in Quyang and Wu (1999) except
the following assumptions: The reorder point r =
expected demand during lead time + SS, and SS
k = (standard deviation of lead time demand),
i.e., r =
*
L +
*
k L o , where
*
= p
1 +
(1 p)
2
,

*
=
2
1 (1 ) p p q + ,
1
=
*
+ (1 p) L qo ,

2
=
*
p L qo (i.e.,
1

2
L qo = , R q e ),
and k is the safety factor which satisfies
( ) 1 p X r p > = u(r
1
) (1 p) u (r
2
) q = , where
u represents the cumulative distribution function
of the standard normal random variable, q repre-
sents the allowable stock-out probability during L,
r
1 =
(r
1
L ) ) ( L o = k
2
1 (1 ) p p q + (1 ) p q ,
and r
2 =
(r
2
L ) ) ( L o = k
2
1 (1 ) p p q + pq + .
Moreover, arrival order may contain some defec-
tive units. In addition, we also assume that the
T h e 2 6 t h W o r k s h o p o n C o m b i n a t o r i a l M a t h e m a t i c s a n d C o m p u t a t i o n T h e o r y
3 1 5
number of non-defective units in an arriving order
of size Q be a binomial random variable with pa-
rameters Q and 1 u , where u (0 u s s 1) repre-
sents the probability of defect. Upon arrival of
order, the entire items are inspected and defective
units in each lot will be returned to the vendor at
the time of delivery of the next lot.
3. Model formulation
3.1 The mixtures of normal distribution
model
First we consider the lead time demand X has a
mixture of normal distribution F
*
= pF
1
+ (1 p)F
2
,
where F
1
has a normal distribution with finite
mean
1
and standard deviation L o and F
2
has
a normal distribution with finite mean
2
and
standard deviation L o ,
1

2
L qo = , R q e .
Therefore, the demand of the lead time X has
mixtures of probability density function (p.d.f.)
which is given by
2
1
2
2
1
2
1
2
( )
( )
1
( )
2
1
(1 )
2
x L L
x L L
f x p e
L
p e
L
o
o
to
to

(

(

=
+
,
where
1

2
L qo = , R q e , xeR, 0 s p s 1,
0 o > (see Everitt and Hand, 1981).
Moreover, if (
1

2
)
2
<
2
27 (8 ) L o or
27/8 q < for any 0 s p s 1, then the mixture of
normal distribution is a unimodal distribution.
When (
1

2
)
2
>
2
4 L o or 2 q > , at least we
can find a (0 1) p p s s makes the mixtures of
normal distribution to be a bimodal distribution.
As mentioned earlier, we have assumed that
shortages are allowed and the reorder point
* *
r L k L o = + , where k,
*
and
*
o are de-
fined as above. Then the expected demand short-
age at the end of the cycle is given by
*
( ) [ ] ( ) ( )
r
B r E X r x r dF x

+
= =
}
1 2
( , , ) , L r r p o = + (1)
where + (r
1
,r
2
,p) = p[ | (r
1
) r
1
(1 u (r
1
))] +
(1 p)[ | (r
2
) r
2
(1 u (r
2
) ) ], | and u be
the standard normal p.d.f. and cumulative distribu-
tion function (c.d.f.), respectively. Thus, the ex-
pected number of backorders per cycle is ( ) B r |
and the expected lost sales per cycle is
(1 ) ( ) B r | , where (0 1) | | s s is the fraction
of the demand during the stock-out period being
backorder. Hence, the stock-out cost per cycle is
0
[ (1 )] ( ) B r t t | + .
Then, the expected net inventory level just be-
fore the order arrives is
)|
( 0 )
* *
1
*
2
*
[( ) ] ( )
(1 )
(1 ) (1 )
(1 ) ( )
X r
E X r I B r
L L
L p r p
L
p p r p
L
p B r
|

o q |
o o

q q |
o

q |
o

< <
+

= u +
+ u
+
| | |
|
\ . \
| |
|
\ .
| |
(
` |
(
)
\ .
and the expected net inventory level at the begin-
ning of the cycle, given that there are y
non-defective items in an arriving order of size Q,
is
)|
( 0 )
* *
1
*
2
*
[( ) ] ( )
(1 )
(1 ) (1 )
(1 ) ( )
X r
y E X r I B r
L L
y L p r p
L
p p r p
L
p B r
|

o q |
o o

q q |
o

q |
o

< <
+
+
= + u +
+ u
+
| | |
|
\ . \
| |
|
\ .
| |
(
` |
(
)
\ .
Therefore, the expected holding cost per cycle
is
)
*
1
* *
2
*
(1 )
2
(1 ) (1 )
(1 ) ( )
L y y
h L p r p
D
L L
p p r
L
p p B r

o q
o

| q
o o

q | q |
o
+ u +
+ + u
+
| |
|
\ .
| |( |
| (
\ . \
| |(
` ` |
(
) ) \ .
If we let L
0
=
1
n
j=
b
j
and L
i
be the length of lead
time with components 1,2,, i crashed to their
minimum duration, then L
i
can be expressed as L
i
=
1
n
j=
b
j

1
i
j=
(b
j
a
j
), i = 1,2, ... ,n; and the
lead time crashing cost R(L) per cycle for a given
that there are y non-defective items in an arriving
order size Q is
( ) C y = ordering cost + non-defective holding
cost + stock-out cost + defective holding cost
+ inspecting cost + lead time crashing cost
T h e 2 6 t h W o r k s h o p o n C o m b i n a t o r i a l M a t h e m a t i c s a n d C o m p u t a t i o n T h e o r y
3 1 6
*
1
* *
2
*
0
(1 )
2
(1 ) (1 )
(1 ) ( ) [
(1 )] ( ) ( ) ( ).
L y y
A h L p r p
D
L L
p p r
L
p p B r
y
B r h Q y vQ R L
D

o q
o

| q
o o

q | q | t
o
t |
= + + u +
+ + u
+ +
'
+ + + +
| |
|
\ .
| |( |
| (
\ . \
| |(
` ` |(
\ . ) )
(2)
The expected length of the cycle time and cycle
cost under the lot of size Q are, respectively,
( | )
( | )
E Y Q
E T Q
D
= (3)
and
* *
1
* *
2
2
0
2
(1 ) (1 )
(1 )
( | ) ( | ) ( | )
2
(1 ) ( ) ( | ) [ (1 )] ( )
( | ) ( | ) ( ).
L L
p r p p
L L
p r p p
h h L
E C Q A E Y Q E Y Q
D D
h
B r E Y Q B r
D
h Q h
E Y Q E Y Q vQ R L
D D

q | q
o o

q | q
o o
o
| t t |
u + +
+ u
= + +
+ + +
' '
+ + +
| | | |(

| | (
\ . \ .
| | | |(
`
| | (
\ . \ . )
(4)
Hence, the expected total annual cost is
)|
* *
1
*
2
*
2
2
0
(1 )
( | ) ( | )
( , )
( | ) ( | ) 2 ( | )
(1 ) (1 )
(1 ) ( ) [
( ) ( | )
(1 )]
( | ) ( | )
N
L L
r p
L
r p
L
p
E C Q AD h E Y Q
EAC Q L
E T Q E Y Q E Y Q
h L p
p p
h B r
DB r E Y Q
h Q h
E Y Q E Y Q

q |
o o

q
o

| q
o
o
q
| t
t |
u +
u

= = +
+
+ +
+ +
' '
+ +
| | |

|
\ . \
| |
|
\ .
( | |
`
|(
) \ .
( ).
( | ) ( | )
DvQ D
R L
E Y Q E Y Q
+ + (5)
For a given lot of size Q, we assume that the
number of non-defective units is a random vari-
able (Y), which has a binomial distribution with
parameter Q and 1 u . That is, Y has the bino-
mial p.d.f. as
( ) (1 ) ,
Q y Q y
y
p y C u u

= (6)
where y = 0,1,2, Q, and 0 u s s 1.
In this case,
( | ) (1 ) E Y Q Q u = (7)
and
2
( | ) (1 )[ (1 )]. E Y Q Q Q u u u = + (8)
Substituting Eqs. (1), (7) and (8) into Eq. (5),
we get
*
1
*
* *
2
1 2 0
1 2
( , ) [ (1 )]
(1 ) 2
(1 )
(1 ) (1 )
(1 ) ( , , ) [ (1 )]
( , , )
( 1)
(1 ) 1
N
AD h
EAC Q L Q
Q
L
h L p r p
L
p p
L L
r p p
h L r r p
D L r r p Dv
h Q
Q
u u
u

o q
o

| q
o

q | q
o o
| o t t |
o
u
u u
= + +

+ u +
+ +
u
+ + + +
+
'
+ +

| |
|
\ .
| |(
|(
\ .
| | | |(
` | | (
\ . \ . )
( ).
(1 )
D
R L
Q u
+

(9)
The parameter (0 1) | s s is treated as a con-
stant; however, in the real market as unsatisfied
demands occur, the longer the length of lead time
is, the larger the amount of shortages is, the
smaller the proportion of customers can wait, and
hence the smaller the backorder rate would be.
Therefore, in model (9), we consider the backor-
der rate | , as a decision variable instead of the
constant case to accommodate the practical in-
ventory situation. During the stock-out period, the
backorder rate | , is variable and is a function of
L through ( ) [ ] B r E X r
+
= . The larger the ex-
pected shortage quantity is, the smaller the back-
order rate would be. Thus we define
, 0 1 , 0 .
1 ( ) B r
o
| o c
c
= s s s <
+
(10)
As the value of c increases, the expected total
annual cost becomes close to the complete lost
case (i.e. 0 | ) for 1 o = . Conversely, decreas-
ing the value of c , the expected total annual cost
T h e 2 6 t h W o r k s h o p o n C o m b i n a t o r i a l M a t h e m a t i c s a n d C o m p u t a t i o n T h e o r y
3 1 7
will approach the complete backordered case
(i.e. 1 | ) for 1 o = .
By using Eq. (10), the model (9) is the expected
annual cost of our new model reduced to
)|
0, 0
* *
1
*
2
*
0
Min EAC ( , ) [ (1 ) ]
(1 ) 2
(1 )
(1 ) (1 )
1 ( )
1 ( )
1 ( )
(1 ) 1 ( )
N
Q L
AD h
Q L Q
Q
L L
h L p r p
L
p p r p
L
p h B r
B r
D
B r
Q B r
u u
u

o q |
o o

q q
o
o
| q
o c
o
t t
u c
> >
= + +

+ u +
+ + u
+
+
+ +
+
| | |
|
\ . \
| |
|
\ .
| |(
| |
` | | (
\ .
\ . )
| |(
|
(
\ .
( 1) ( ) ,
1 (1 )
Dv D
h Q R L
Q
u
u u
'
+ + +

(11)
In order to find the minimum total expected
annual cost, taking the first partial derivatives of
( , )
N
EAC Q L with respect to Q and L in each time
interval [L
i
, L
i-1
], we obtain
2
0 2
EAC ( , )
(1 )
(1 ) 2
1 ( )
(1 ) 1 ( )
N
Q L AD h
Q Q
D
B r
Q B r
u
u
o
t t
u c
c
= +
c
+
+
| |(
|
(
\ .
2
( ).
(1 )
D
h R L
Q
u
u
'
+

(12)
and
)
)
*
1
* *
2
* * *
1
*
* *
2
EAC ( , )
(1 )
2
(1 ) (1 )
2
(1 ) (1 ) (1 )
(
N
Q L h L
p r p
L L
L L
p p r
L h L
p p p r
L
p p p
L L hB r
r p p
o
q
o

| q
o o

q | q
o o

q | q
o

q | q
o o
c
= u +
c
+ + u
+ +
+ + +
+ +
| |
|
\ .
| |( |
| (
\ . \
|
| |(
` |(
) \ . \
| |
|
\ .
| | | |

` | |
\ . \ . )
)
2L
2
0
0 2
( ) ( )
1
1 ( ) 2 1 ( ) 2 (1 )
( )
1
[1 ( )] 1 ( )
h B r DB r
B r L B r Q L
B r
B r B r
o oc
c c u
t oc o
t t
c c
+ +
+ +
+ +
+ +
| | | |
| |
\ . \ .
| |(
` |
(
\ . )
,
(1 )
i
D
c
Q u

(13)
where ( ) B r L o = + (r
1
, r
2
, p).
It is clear that for any given r
1
, r
2
and p, we
have + (r
1
, r
2
, p) >0. Hence, for fixed L e[ L
i
,
L
i-1
] we obtain
| {
2
0 2 3
EAC ( , ) 2
(1 )
1 ( ) ( ) 0
1 ( )
N
Q L D
A
Q Q
B r R L
B r
t t
u
o
c
c
= + +
c
+ >
+
| |(
` |
(
\ . )
then ( , )
N
EAC Q L is convex in Q. However, we
can also show that for fixed Q, ( , )
N
EAC Q L is
concave function of L e[ L
i
, L
i-1
], if
* *
min (1 ) , 2
L L
p p

q q
o o
+ >

`
)
.
Therefore, for fixed Q, the minimum expected
total annual cost will occur at the end points of the
interval [ L
i
, L
i-1
]. Setting Eq. (12) equal to zero
and solving for Q, we have
| |
| {
0
2
(1 ) (1 ) 2
D
Q A
h h
t t
u u u
= + +
'
+

1
2
1 ( ) ( )
1 ( )
B r R L
B r
o
c
+
+
| |(
`` |
(
\ . ))
(14)
where ( ) B r and ( ) R L as the above definitions.
Thus, we can establish the following iterative al-
gorithm to find the optimal lead time and the op-
timal order quantity.
Algorithm 1
Step 1 For given A, D, h, ' h , v, , ,
0
, , p,
q, , , ,
i
a , b
i
, c
i
, and i = 1,2, ,n.
Step 2Use the
i
a , b
i
and c
i
, i = 1,2, ,n,
compute L
i
and compute B(r) using Eq. (1).
Step 3Given , p and q by using the computer
software Intel Visual Fortran V9.0 (Inclusive of
IMSL) (2005) and the subroutine ZREAL from
IMSL to solve k of the equation 1 p(r
1
) (1 p)
(r
2
) = q where r
1
= k
2
1 (1 ) p p q + (1 p)
and r
2
= k
2
1 (1 ) p p q + + p. Further, we obtain
r
1
and r
2
.
Step 4For each L
i
, i = 1,2, ,n, compute Q
i
by
T h e 2 6 t h W o r k s h o p o n C o m b i n a t o r i a l M a t h e m a t i c s a n d C o m p u t a t i o n T h e o r y
3 1 8
using Eq. (14).
Step 5For each pair (Q
i
, L
i
), compute the cor-
responding expected total annual cost EAC
N
(Q
i
,
L
i
), i = 1,2, ,n.
Step 6Find min EAC
N
(Q
i
, L
i
). If EAC
N
(Q
*
,
L
*
) = min EAC
N
(Q
i
, L
i
), then (Q
*
, L
*
) is the opti-
mal solution.
3.2 The mixtures of distribution free model
The information about the form of the mixtures
of d.f. of lead time demand is often limited in
practices. In this subsection, we relax the restric-
tion about the form of the mixtures of d.f. of lead
time demand, i.e., we assume here that the lead
time demand X has a mixture of d.f.
F
*
= pF
1
+ (1 p)F
2
, where F
1
has finite mean
1
and standard deviation L o and F
2
has a finite
mean
2
and standard deviation L o ,

2 =

L o
, eR, but make no assumption
on the mixtures of d.f.s form of F
*
. Now, we try
to use a minimax distribution free procedure to
solve this problem. If we let O denote the class
of all single c.d.f. (include F
1
and F
2
) with finite
mean and standard deviation then the minimax
distribution free approach for our problem is to
find the most unfavorable c.d.f.s F
1
and F
2
in O
for each decision variable and then to minimize
over the decision variables; that is, our problem is
to solve
1 2
0, 0 ,
min max ( , )
F
Q L F F
EAC Q L
> > eO
(15)
where
( ) | |
( ) | |
*
0
( , ) 1
(1 ) 2
(1 ) ( )
1 ( )
(1 )
( 1)
1
( ) ,
(1 )
F
AD h
EAC Q L Q
Q
h k L B r
D
B r
Q
Dv
h Q
D
R L
Q
u u
u
o |
t t |
u
u
u
u
= + +

+ +
+ +

'
+ +

(

and ( ) ( ) .
1 ( )
B r E X r
B r
o
|
c
+
= =
+

`
)
In addition, we need the following Proposition 1
to solve the above problem of the model (15).
Proposition 1. For F
1
and F
2
eO ,
{ }
2 2
1
( ) ( ) ( )
2
i i i
E X r L r L r L o
+
s +
i = 1,2, (16)
where r.v.s X
i
has a single distribution function F
i
,
i = 1,2. Moreover, the upper bound (16) is tight.
So, by using F
*
= pF
1
+ (1 p)F
2
and Eq. (16),
we obtain
( )
2
2
2
2
2
1
( ) 1 (1 )
2 2
1 1 (1 ) (1 )
1
1 1 (1 )
2
p
B r k p p L L
k p p p
p
L k p p p
q o o
q q
o q q
s + +
+ +

+ + + +

(
`

)

(
`

)
(17)
From the definition of the backorder rate | and
inequality (17), we have
( )
{ {
2
2
2
1
2
2
1 ( )
1
1 1 (1 )
2
1 1 (1 ) (1 )
2
1
1 1 (1 )
2
B r
k p p L
p
L k p p p
p
L k p p p
o
|
c
o c q o
o q q
o q q

=
+
> + +
+ + +

+ + + +

`
)
(
(
(


(
(
``
(

))
(18)
where 0 1 , 0 , 0 1 , 0. p o c o s s s s s s >
By using the inequality (17) and (18), the model
(15) can be reduced to
0, 0
min ( , )
U
Q L
EAC Q L
> >
(19)
where
2
0
(1 )
( , )
(1 ) 2
1 (1 )
(1 )
( )
1
(1 ) 1 ( )
( 1) ( ) ,
1 (1 )
U
U
U
AD Q
EAC Q L h
Q
D
k p p L
Q
D L
h
Q L
Dv D
h Q R L
Q
u u
u
q o t
u
o
t
u c
u
u u
+
= +

+ + +

A
+ +
+ A
'
+ + +

| | | | | |(
| | |
(
\ . \ . \ .
u = > = = = > = > > ' = s = > > = u

{
2
2
2
2
2
1
( ) 1 (1 )
2
1 1 (1 ) (1 )
2
1
1 1 (1 )
2
U
L k p p L
p
L k p p p
p
L k p p p
c q o
o q q
o q q
A = +
+ + +

+ + + +
(

(
(
(


(
(
`
(

)
and EAC
U
(Q,L) and
U
(L) are the least upper
bounds of EAC
F
(Q,L) and (L), respectively.
To minimize Eq. (19), we take the partial de-
rivatives of EAC
U
(Q,L) with respect to Q and L,
we obtain
2
0
2
( , )
(1 )
(1 ) 2
( )
1
(1 ) 1 ( )
U
U
U
EAC Q L AD h
Q Q
D L
Q L
u
u
o
t t
u c
c
= +
c
A
+
+ A
| | | |(
| |
(
\ . \ .
2
( )
(1 )
D
h R L
Q
u
u
'
+

(20)
and
2
2
0
2
0
( , )
1 (1 )
2
[ ( )]
2 (1 ) [1 ( )]
( )
1
1 ( ) 2
( )
1
(1 ) 1 ( ) 2
U
U
U
U
U
U
U
EAC Q L hk
p p
L L
D L
h
Q L L
L
h
L L
D L
Q L L
q o
o
t
c u
o
c
o
t t
u c
c
= +
c
A
+ +
+ A
A
+
+ A
A
+ +
+ A
(

| | | |
| |
\ .\ .
| | | |
| |
\ .\ .
| | | |(
| |
(
\ . \ .
(1 )
i
D
c
Q u

, (21)
where
U
(L) = L o + (r
1
, r
2
, p) = B(r).
In addition, it can be verified that for fixed L
e[ L
i
, L
i-1
], EAC
U
(Q,L) is convex in Q. For fixed
Q, EAC
U
( Q , L ) is concave in L e[ L
i
, L
i-1
].
Therefore, for fixed Q, the minimum expected
total annual cost will occur at the end points of the
interval L e[ L
i
, L
i-1
]. Hence, setting Eq.(20)
equal to zero and solving for Q, we get
| |
| {
1
2
0
2
(1 ) (1 ) 2
( )
1 ( )
1 ( )
U
U
D
Q A
h h
L
R L
L
t
u u u
o
t
c
= +
'
+
A
+ +
+ A

| | | |(
`` | |
(
\ .\ . ))
(22)
Theoretically, for fixed A, D, h, ' h , v, , ,
0
,
, p, q, , , , and each L
i
(i = 1,2, ,n), the opti-
mal (Q
i
, L
i
) pair given L
i
can be obtained by solv-
ing Eq. (22) iteratively until convergence. The
convergence of the procedure can be shown. Fur-
thermore, using Eq. (19), we can obtain the corre-
sponding expected total annual cost EAC
U
(Q
i
, L
i
).
Hence, the minimum expected total annual cost is
min
i = 0,1, ,n
EAC
U
(Q
i
, L
i
). However, in practice,
since the p.d.f.
X
of X is unknown, even if the
value of q is given, we can not get the exact value
of k. Thus, in order to find the value of k, we need
the following proposition.
Proposition 2. Let Y be a random variable
which has a p.d.f.
Y
(y) with finite mean L and
standard deviation ( 0) L o > , the for any real
number d > L,
2
2 2
( ) .
( )
L
P Y d
L d L
o
o
> s
+
(23)
So, by using F
*
= pF
1
+ (1 p)F
2
, the recorder
point r =
*
L +
*
k L o and the Proposition 2, we
get
2
2
2
2
( )
1 1 (1 ) (1 )
1
1 1 (1 )
p
P X r
k p p p
p
k p p p
q q
q q
> s
+ +

+
+ + +
(

(

(24)
Further, it is assumed that the allowable
stock-out probability q during lead time is given,
that is, ( ) q P X r = > , then from Eq. (24), we get
0 1 1 k q q s s + .
It is easy to verify that EAC
U
(Q,L) has a smooth
curve for 0 1 1 k q q s s + . Hence, we can es-
tablish the following algorithm to obtain the suit-
able k and hence the optimal Q and L.
Algorithm 2
Step 1 For given A, D, h, ' h , v, , ,
0
, ,
p, q, , , ,
i
a , b
i
, c
i
, and i = 1,2, ,n.
Step 2For a given q, we divide the interval
[0, 1 1 ] q q + into m equal subintervals, where
m is large enough. And we let k
0
= 0,
k
m
1 1 q q = + and k
j
= k
j-1
+ ( k
m
k
0
)/m,
j = 1,2, , m 1.
Step 3Use the
i
a , b
i
and c
i
, to compute L
i
,
i = 1,2, ,n.
Step 4For each L
i
, i = 1,2, ,n, compute Q
i, kj
by using Eq. (22) for given k
j
e{ k
0
, k
1
, , k
m
},
j = 0,2, , m
Step 5For each L
i
, i = 1,2, ,n, compute the
corresponding expected total annual cost
T h e 2 6 t h W o r k s h o p o n C o m b i n a t o r i a l M a t h e m a t i c s a n d C o m p u t a t i o n T h e o r y
3 2 0
EAC
U
(Q
i, kj
, L
i
),.
Step 6Find min
kj e{ k0, k1, , km}
EAC
U
(Q
i, kj
, L
i
)
and let EAC
U
(Q
i, ks(i)
, L
i
) = min
kj e{ k0, k1, , km}
EAC
U
(Q
i, kj
, L
i
), then find min
i = 1,2, ,n
EAC
U
(Q
i,
ks(i)
, L
i
). If EAC
U
(Q
*
, L
*
) = min
i = 1,2, ,n
EAC
U
(Q
i,
ks(i)
, L
i
), then (Q
*
, L
*
) is the optimal solution.
4. Numerical examples
In order to illustrate the above solution proce-
dure and the effects of ordering cost reduction, let
us consider an inventory system with the data: D
= 600 units year , A = $200 per order, h = $20,
' h = $12, v = $1.6, = $50,
0
= $150,
*
= 11
units week , o = 7 units week , u = 0.00, 0.15,
0.30, 0.45, c = 0, 2, 100, (backorder case), q =
0.2 and the lead time has three components with
shown in Table 1.
Example 1. We assume here that the lead
time demand follows a mixtures of normal distri-
bution and want to solve the case when 0.7 q = ,
p = 0, 0.4, 0.8, 1.0, c = 0, 2, 100, , u = 0.00,
0.15, 0.30, 0.45 and o = 1.0, 0.6. A summary of
these optimal results is presented in (Q
*
, L
*
) and
EAC
N
(Q
*
, L
*
) of Tables 2 - 3. From Tables 2 - 3,
when 0.7 q = , for the fixed p, defective rate u
and o = 0.6, 1.0, we can get the order quantity Q
*
and the minimum expected total annual cost
EAC
N
(Q
*
, L
*
) decrease as c decreases ( i.e. the
backorder rate | increases ) (also see Fig. 2).
Moreover, for the fixed p, c and o = 0.6, 1.0,
the order quantity Q
*
and the minimum expected
total annual cost EAC
N
(Q
*
, L
*
) increase as u
increases (also see Fig. 1); for o = 0.6, 1.0, the
fixed c and defective rate u , when p = 0 or 1,
the model considers only one kind of customers
demand; when 0 p < < 1, the model considers two
kinds of customers demand. It implies that
EAC
N
(Q
*
, L
*
) of two kinds of customers demand
are larger than EAC
N
(Q
*
, L
*
) of one kind of cus-
tomers demand. Thus, the minimum expected
total annual cost EAC
N
(Q
*
, L
*
) increases as the
distance between p and 0 (or1) increases for the
fixed c and defective rate u (also see Fig. 3).
In addition, no matter what values of p andu , the
optimal lead time L
*
is approached to a certain
value (3 weeks) for o = 0.6, 1.0 and c = 2,
100, . However, no matter what values of p
andu , the optimal lead time L
*
is approached to a
certain value (4 weeks) for o = 0.6, 1.0 and
c = 0.
------------- Insert Tables 2 - 3 here ------------
-------------- Insert Figs. 1 - 3 here -------------
Example 2. The data is as in Example 1. We
assume here that the probability distribution of the
lead time demand is mixtures of free. By using the
proposed Algorithm 2 and m = 500 procedure
yields these optimal results shown in (Q
*
, L
*
) and
EAC
U
(Q
*
, L
*
) of Tables 2 - 3. From Tables 2 - 3,
when 0.7 q = , for the fixed p , defective rate u
and o = 0.6, 1.0, we can get the order quantity
Q
*
and the minimum expected total annual cost
EAC
U
(Q
*
, L
*
) decrease as c decreases ( i.e. the
backorder rate | increases ) (also see Fig. 2).
Moreover, for the fixed p , c and o = 0.6, 1.0,
the order quantity Q
*
and the minimum expected
total annual cost EAC
U
(Q
*
, L
*
) increase as u
increases (also see Fig. 1); the minimum expected
total annual cost EAC
U
(Q
*
, L
*
) increases and then
decreases as p increases for o = 0.6, 1.0 and
the fixed c and 0, 0.15 u = (also see Fig. 3). In
addition, no matter what values of p andu , the
optimal lead time L
*
is approached to a certain
value (3 weeks) for o = 0.6, 1.0 and c = 2,
100, . However, no matter what values of
p andu , the optimal lead time L
*
is approached to
a certain value (3 weeks) for o = 0.6 and c = 0
and L
*
is approached to a certain value (4 weeks)
for o = 1.0 and c = 0. Finally, the expected total
annual cost EAC
N
(Q
*
, L
*
) is obtained by substitut-
ing Q
*
and L
*
into Eq. (11) when the d.f. of the
lead time demand is the mixtures of normal dis-
tribution. The expected value of additional infor-
mation (EVAI) is the largest amount that one is
willing to pay for the knowledge of F
1
and F
2
and
is equal to EAC
N
(Q
*
, L
*
) EAC
N
(Q
*
, L
*
). From
Tables 2 - 3, we observe that for the fixed p
andu , EVAI decreases as c increases (i.e the
backorder rate | decreases) except c = 0, when
o = 0.6, 1.0. Moreover, we also observe that for
the given c = 2, 100, and fixed u , EVAI in-
creases and then decreases as p increases when
o = 0.6, 1.0. And the cost penalty EAC
N
(Q
*
, L
*
)/
EAC
N
(Q
*
, L
*
) of using the distribution free oper-
ating policy instead of the optimal one is decreas-
ing as c increases (i.e the backorder rate |
decreases) except c = 0 fixed p andu .
5. Concluding remarks
In this article, we propose a continuous review
inventory model by considering the mixtures of
distribution of the lead time demand with defec-
Table 1 Lead time data
Lead time
component
i
Normal
duration
bi(days)
Minimum
duration
i
a (days)
Unit crash-
ing cost
ci
($ days)
1
2
3
20
20
16
6
6
9
0.4
1.2
5.0
T h e 2 6 t h W o r k s h o p o n C o m b i n a t o r i a l M a t h e m a t i c s a n d C o m p u t a t i o n T h e o r y
3 2 1
tive units and controllable backorder rate to extend
that of Ouyang and Chuang (2001) (also see Lee
(2005)). In addition, we also develop an algo-
rithmic procedure to find the optimal lead time.
Further, we get the significant results in the ex-
pected total annual cost.
In future research on this problem, it would be
interesting to deal with the inventory model with a
service level constraint or treat the reorder point as
a decision variable.
Acknowledgements
This research was partially supported by the
National Science Council, ROC (Plan No.
NSC97-2221-E-309-008).
References
[1] Ben-Daya, M. and Raouf, A., 1994, Inventory
models involving lead time as decision vari-
able, Journal of the Operational Research So-
ciety, 45, 579-582.
[2] Everitt, B. S. and Hand, D. J., 1981, Finite
Mixture Distribution, Chapman & Hall, Lon-
don, NewYork.
[3] Gallego, G. and Moon. I., 1993, The distribu-
tion free newsboy problem: Review and ex-
tensions, Journal of the Operational Research
Society, 44, 825-834.
[4] Intel Visual Fortran (Inclusive of IMSL),
Standard Edition for Windows Version 9,
Copy- right(c), Intel Corporation, 2005.
[5] 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,
701-717.
[6] Liao, C. J. and Shyu, C. H., 1991, An analyti-
cal determination of lead time with normal
demand, International Journal of the Opera-
tions Production Management, 11, 72-78.
[7] Naddor, E., 1996, Inventory Systems,
John-Wiley, New York.
[8] Ouyang, L. Y., Yeh, N. C. and Wu, K. S.,
1996, Mixture inventory model with backor-
ders and lost sales of variable lead time,
Journal of the Operational Research Society,
47, 829-832.
[9] Ouyang, L. Y. and Wu, K. S., 1999, Mixture
inventory model involving variable lead time
with defective units, Journal of Statistic and
Management System, 2, 143-157.
[10] Ouyang, L. Y., Chuang, B. R. and Wu, K. S.,
1999, Optimal inventory policies involving
variable lead time with defective items, Op-
search, 36, 374-389.
[11] Ouyang, L. Y. and Chuang, B. R., 2001, Mix-
ture inventory model involving variable lead
time and controllable backorder rate, Comput.
Ind. Eng., 40, 339-348.
[12] Silver, E. A., Peterson, R., 1985, Decision
Systems for Inventory Management and Pro-
duction Plannings, John-Wiley, New York.
[13] Tersine, R. J., 1982, Principles of Inventory
and Materials Management, North-Holland,
NewYork.
[14] Wu, J. W. and Tsai, H. Y., 2001, Mixture in-
ventory model with backorders and lost sales
for variable lead time demand with the mix-
tures of normal distribution, International
Journal of Systems Science, 32, 259-268.
T h e 2 6 t h W o r k s h o p o n C o m b i n a t o r i a l M a t h e m a t i c s a n d C o m p u t a t i o n T h e o r y
3 2 2
Table 2 Summary of the optimal solution procedure ( L
i
in weeks and 0.7, 1.0 q o = = )
( *, *) ( *, *) ( *, *) ( , ) ( , ) EVAI Cost penalty
* * * *
U N N
Q L EAC Q L EAC Q L Q L EAC Q L c u
p = 0.0 ( or 1 )

0.00 (169, 3) 5011.812 4754.263 (178, 3) 4749.207 5.057 1.0011

0.15 (180, 3) 5521.167 5282.753 (190, 3) 5277.187 5.566 1.0011
0.30 (196, 3) 6200.876 5987.094 (207, 3) 5980.871 6.223 1.0010
0.45 (218, 3) 7172.743 6992.599 (230, 3) 6985.479 7.119 1.0010
100
0.00 (169, 3) 5006.326 4749.065 (178, 3) 4743.986 5.080 1.0011
0.15 (180, 3) 5515.148 5277.051 (190, 3) 5271.459 5.592 1.0011
0.30 (196, 3) 6194.171 5980.743 (206, 3) 5974.493 6.250 1.0010
0.45 (218, 3) 7165.101 6985.362 (230, 3) 6978.211 7.151 1.0010
2
0.00 (159, 3) 4817.129 4557.497 (169, 3) 4552.247 5.250 1.0012
0.15 (170, 3) 5307.551 5066.889 (180, 3) 5061.109 5.779 1.0011
0.30 (185, 3) 5962.887 5746.644 (196, 3) 5740.184 6.460 1.0011
0.45 (206, 3) 6901.468 6718.575 (218, 3) 6711.184 7.391 1.0011
0
0.00 (142, 4) 4185.610 3884.810 (134, 4) 3880.775 4.035 1.0010
0.15 (150, 4) 4639.367 4323.690 (144, 4) 4320.613 3.077 1.0007
0.30 (161, 4) 5243.830 4912.849 (156, 4) 4910.833 2.016 1.0004
0.45 (177, 4) 6107.398 5761.479 (173, 4) 5760.489 0.990 1.0002
p = 0.4

0.00 (167, 3) 5018.595 4828.015 (181, 3) 4817.125 10.890 1.0023

0.15 (179, 3) 5524.946 5362.674 (193, 3) 5350.686 11.988 1.0022
0.30 (194, 3) 6200.787 6074.957 (210, 3) 6061.555 13.401 1.0022
0.45 (216, 3) 7167.371 7091.307 (233, 3) 7075.976 15.331 1.0022
100
0.00 (167, 3) 5013.064 4822.918 (181, 3) 4811.979 10.939 1.0023
0.15 (178, 3) 5518.877 5357.083 (193, 3) 5345.042 12.042 1.0023
0.30 (194, 3) 6194.026 6068.730 (210, 3) 6055.269 13.461 1.0022
0.45 (216, 3) 7159.666 7084.212 (233, 3) 7068.812 15.400 1.0022
2
0.00 (158, 3) 4824.652 4631.414 (171, 3) 4620.086 11.328 1.0025
0.15 (169, 3) 5312.134 5147.003 (183, 3) 5134.533 12.470 1.0024
0.30 (183, 3) 5963.687 5834.737 (199, 3) 5820.797 13.940 1.0024
0.45 (204, 3) 6897.101 6817.562 (221, 3) 6801.615 15.948 1.0023
0
0.00 (142, 4) 4203.567 3918.194 (135, 4) 3915.470 2.723 1.0007
0.15 (150, 4) 4656.681 4359.345 (145, 4) 4357.452 1.894 1.0004
0.30 (161, 4) 5260.390 4951.497 (157, 4) 4950.432 1.065 1.0002
0.45 (177, 4) 6123.053 5804.198 (175, 4) 5803.856 0.342 1.0001
p = 0.8

0.00 (168, 3) 5015.096 4787.362 (179, 3) 4779.563 7.799 1.0016

0.15 (179, 3) 5522.304 5318.413 (191, 3) 5309.827 8.585 1.0016
0.30 (195, 3) 6199.250 6026.050 (208, 3) 6016.453 9.597 1.0016
0.45 (217, 3) 7167.342 7036.057 (231, 3) 7025.078 10.979 1.0016
100
0.00 (167, 3) 5009.578 4782.207 (179, 3) 4774.373 7.834 1.0016
0.15 (179, 3) 5516.250 5312.758 (191, 3) 5304.135 8.624 1.0016
0.30 (194, 3) 6192.505 6019.752 (208, 3) 6010.112 9.640 1.0016
0.45 (216, 3) 7159.654 7028.881 (231, 3) 7017.853 11.028 1.0016
2
0.00 (158, 3) 4820.929 4590.640 (170, 3) 4582.543 8.096 1.0018
0.15 (169, 3) 5309.249 5102.601 (181, 3) 5093.689 8.912 1.0017
0.30 (184, 3) 5961.880 5785.666 (197, 3) 5775.703 9.962 1.0017
0.45 (204, 3) 6896.768 6762.114 (219, 3) 6750.717 11.397 1.0017
0
0.00 (142, 4) 4195.115 3901.879 (135, 4) 3898.406 3.473 1.0009
0.15 (150, 4) 4648.576 4341.640 (144, 4) 4339.120 2.520 1.0006
0.30 (161, 4) 5252.694 4932.020 (156, 4) 4930.467 1.553 1.0003
0.45 (177, 4) 6115.846 5782.344 (174, 4) 5781.662 0.682 1.0001
T h e 2 6 t h W o r k s h o p o n C o m b i n a t o r i a l M a t h e m a t i c s a n d C o m p u t a t i o n T h e o r y
3 2 3
Table 3 Summary of the optimal solution procedure ( L
i
in weeks and 0.7, 0.6 q o = = )
( *, *) ( *, *) ( *, *) ( , ) ( , ) EVAI Cost penalty
* * * *
U N N
Q L EAC Q L EAC Q L Q L EAC Q L c u
p = 0.0 ( or 1 )

0.00 (169, 3) 5011.812 4754.264 (178, 3) 4749.207 5.057 1.0011

0.15 (180, 3) 5521.167 5282.753 (190, 3) 5277.187 5.567 1.0011
0.30 (196, 3) 6200.876 5987.094 (207, 3) 5980.871 6.223 1.0010
0.45 (218, 3) 7172.743 6992.599 (230, 3) 6985.480 7.119 1.0010
100
0.00 (169, 3) 5008.521 4751.146 (178, 3) 4746.075 5.070 1.0011
0.15 (180, 3) 5517.557 5279.333 (190, 3) 5273.751 5.582 1.0011
0.30 (196, 3) 6196.854 5983.285 (206, 3) 5977.045 6.240 1.0010
0.45 (218, 3) 7168.160 6988.258 (230, 3) 6981.120 7.138 1.0010
2
0.00 (163, 3) 4896.300 4637.449 (172, 3) 4632.283 5.166 1.0011
0.15 (174, 3) 5394.426 5154.605 (184, 3) 5148.919 5.686 1.0011
0.30 (189, 3) 6059.679 5844.356 (200, 3) 5838.000 6.356 1.0011
0.45 (211, 3) 7011.805 6829.937 (223, 3) 6822.666 7.271 1.0011
0
0.00 (156, 3) 4609.345 4295.486 (154, 4) 4283.872 11.614 1.0027
0.15 (165, 3) 5090.929 4779.604 (164, 4) 4763.082 16.522 1.0035
0.30 (177, 3) 5730.780 5427.273 (179, 4) 5403.986 23.287 1.0043
0.45 (195, 3) 6641.947 6356.064 (199, 4) 6322.858 33.206 1.0053
p = 0.4

0.00 (167, 3) 5018.595 4828.015 (181, 3) 4817.125 10.890 1.0023

0.15 (179, 3) 5524.946 5362.674 (193, 3) 5350.686 11.988 1.0022
0.30 (194, 3) 6200.787 6074.957 (210, 3) 6061.556 13.401 1.0022
0.45 (216, 3) 7167.372 7091.307 (233, 3) 7075.976 15.332 1.0022
100
0.00 (167, 3) 5015.278 4824.958 (181, 3) 4814.039 10.919 1.0023
0.15 (179, 3) 5521.306 5359.321 (193, 3) 5347.300 12.021 1.0022
0.30 (194, 3) 6196.732 6071.222 (210, 3) 6057.785 13.437 1.0022
0.45 (216, 3) 7162.750 7087.051 (233, 3) 7071.679 15.372 1.0022
2
0.00 (162, 3) 4903.529 4711.271 (175, 3) 4700.134 11.137 1.0024
0.15 (173, 3) 5398.690 5234.610 (187, 3) 5222.351 12.259 1.0023
0.30 (188, 3) 6060.127 5932.321 (203, 3) 5918.617 13.704 1.0023
0.45 (209, 3) 7007.040 6928.771 (226, 3) 6913.094 15.678 1.0023
0
0.00 (155, 3) 4621.130 4342.202 (156, 4) 4338.804 3.397 1.0008
0.15 (165, 3) 5102.196 4830.267 (167, 4) 4822.121 8.146 1.0017
0.30 (177, 3) 5741.465 5483.066 (181, 4) 5468.312 14.755 1.0027
0.45 (194, 3) 6651.958 6418.864 (201, 4) 6394.403 24.460 1.0038
p = 0.8

0.00 (168, 3) 5015.096 4787.362 (179, 3) 4779.563 7.799 1.0016

0.15 (179, 3) 5522.304 5318.413 (191, 3) 5309.828 8.585 1.0016
0.30 (195, 3) 6199.250 6026.050 (208, 3) 6016.453 9.597 1.0016
0.45 (217, 3) 7167.342 7036.057 (231, 3) 7025.078 10.979 1.0016
100
0.00 (168, 3) 5011.786 4784.270 (179, 3) 4776.450 7.820 1.0016
0.15 (179, 3) 5518.673 5315.021 (191, 3) 5306.413 8.608 1.0016
0.30 (195, 3) 6195.204 6022.273 (208, 3) 6012.650 9.623 1.0016
0.45 (216, 3) 7162.731 7031.752 (231, 3) 7020.744 11.008 1.0016
2
0.00 (162, 3) 4899.896 4670.562 (174, 3) 4662.596 7.966 1.0017
0.15 (173, 3) 5395.902 5190.283 (185, 3) 5181.514 8.769 1.0017
0.30 (188, 3) 6058.427 5883.336 (201, 3) 5873.534 9.802 1.0017
0.45 (209, 3) 7006.827 6873.427 (224, 3) 6862.213 11.214 1.0016
0
0.00 (156, 3) 4615.837 4317.087 (155, 4) 4309.793 7.294 1.0017
0.15 (165, 3) 5097.175 4802.824 (165, 4) 4790.684 12.140 1.0025
0.30 (177, 3) 5736.749 5452.563 (180, 4) 5433.750 18.814 1.0035
0.45 (195, 3) 6647.590 6384.256 (200, 4) 6355.575 28.681 1.0045
T h e 2 6 t h W o r k s h o p o n C o m b i n a t o r i a l M a t h e m a t i c s a n d C o m p u t a t i o n T h e o r y
3 2 4
3500
4500
5500
6500
7500

=
0
.
0
0

=
0
.
1
5

=
0
.
3
0

=
0
.
4
5

=
1
0
0

=
0
.
0
0

=
0
.
1
5

=
0
.
3
0

=
0
.
4
5

=
2

=
0
.
0
0

=
0
.
1
5

=
0
.
3
0

=
0
.
4
5

=
0

=
0
.
0
0

=
0
.
1
5

=
0
.
3
0

=
0
.
4
5
E
A
C
EACU(**) EACN(**) EACN(*)
Fig. 1. Summary of the results of the optimal procedure based on 0.7 q = , 0.4 p = and 1.0 o = . Note:
* *
( , )
U
EAC Q L ,
* *
( , )
N
EAC Q L and
* *
( , )
N
EAC Q L will be denoted by the symbols EACU(**), EACN(**) and
EACN(*), respectively.
3500
4500
5500
6500
7500

=
0
.
0
0

=
1
0
0

=
2

=
0

=
0
.
1
5

=
1
0
0

=
2

=
0

=
0
.
3
0

=
1
0
0

=
2

=
0

=
0
.
4
5

=
1
0
0

=
2

=
0
E
A
C
EACU(**) EACN(**) EACN(*)
Fig. 2. Summary of the results of the optimal procedure based on 0.7 q = , 0.4 p = and 1.0 o = .
5250
5300
5350
5400
5450
5500
5550
p=0.0 p=0.4 p=0.8 p=1.0
E
A
C
EACU(**) EACN(**) EACN(*)
Fig. 3. Summary of the results of the optimal procedure based on 0.7 q = , 100 c = , 0.15 u = and 1.0 o = . Note:
* *
( , )
U
EAC Q L ,
* *
( , )
N
EAC Q L and
* *
( , )
N
EAC Q L will be denoted by the symbols EACU(**), EACN(**) and
EACN(*), respectively.
T h e 2 6 t h W o r k s h o p o n C o m b i n a t o r i a l M a t h e m a t i c s a n d C o m p u t a t i o n T h e o r y
3 2 5