Solving Solid Transportation Problem With Multi Choice Cost and Stochastic Supply and Demand PDF

International Journal of Strategic Decision Sciences, 5(3), 1-26, July-September 2014 1
Solving Solid Transportation

Problem with Multi-Choice
Cost and Stochastic
Supply and Demand
Sankar Kumar Roy, Department of Applied Mathematics with Oceanology and Computer
Programming, Vidyasagar University, Midnapore, West Bengal, India
Deshabrata Roy Mahapatra, Department of Applied Mathematics with Oceanology and
Computer Programming, Vidyasagar University, Midnapore, West Bengal, India
ABSTRACT
This paper proposes a new approach to analyze the solid transportation problem (STP). This new approach
considers the multi-choice programming into the cost coefficients of objective function and stochastic program-
ming which is incorporated in three constraints namely sources, destinations and capacities constraints fol-
lowed by Cauchys distribution for solid transportation problem. The multi-choice programming and stochastic
programming are combined into solid transportation problem and this new problem is called multi-choice
stochastic solid transportation problem (MCSSTP). The solution concepts behind the MCSSTP are based
on a new transformation technique which will select an appropriate choice from a set of multi-choice which
optimizes the objective function. The stochastic constraints of STP convert into deterministic constraints by
stochastic programming approach. Finally, the authors have constructed a non-linear programming problem
for MCSSTP and have derived an optimal solution of the specified problem. A realistic example on STP is
considered to illustrate the methodology.
Keywords: Cauchys Distribution, Multi-Choice Programming, Solid Transportation Problem, Stochastic

Programming, Transformation Technique
INTRODUCTION each origin to various destinations, satisfying

source availability and destination demand and
The classical transportation problem (TP) minimizing the total cost of transportation. The
can be described as a special case of linear amounts of available goods at the supply points
programming problem. Its model is applied to and the amounts required at the demand points
determine an optimal solution of the TP of how are the parameters of the TP, and these param-
many units of commodity to be shipped from eters are not always exactly known and stable.
DOI: 10.4018/ijsds.2014070101
Copyright 2014, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
2 International Journal of Strategic Decision Sciences, 5(3), 1-26, July-September 2014
This imprecision may follow from the lack of Particularly, in the recent years it has gained
exact information. The STP is an important more importance. The Cauchys distribution
extension of the traditional TP. The traditional is empirical as many large data sets exhibit on
TP is a well known optimization problem in heavy tails and skewness. The strong empirical
operations research, in which two kinds of evidence for these features combined with the
constraints are taken into consideration, i.e., generalized central limit theorem is used for
source constraints and destination constraints. many types of physical and economic systems.
But in the real system, we always deal with In the recent past normal, log-normal,
other constraints besides of source constraints extreme value distribution and other random
and destination constraints, such as product type variables have been considered in the stochas-
constraints or transportation mode constraints. tic programming model. Cauchys distribution
For such case, the traditional TP turns into STP. has also two parameters namely location pa-
The STP is a generalization of the well-known rameter (i ) and scale parameter ( ). It has
TP in which three items (they are source, desti- some similarities with normal distribution. As-
nation and conveyance) are taken into account suming Cauchys distribution is a suitable one
in the constraint set instead of two (source and for stochastic programming model, we con-
destination). In many industrial problems, a ho- sider the source, the availability and the convey-
mogeneous product is delivered from an origin ance parameters of the transportation problem
to destination by means of different modes of as Cauchys distribution. Here, we consider the
transport called conveyances, such as trucks, probability density function of Cauchys dis-
cargo flights, goods trains, ships, etc. In reality,
tribution with location parameter i and scale
due to changes in market supply and demand,
weather conditions, road conditions and other parameter i as:
unpredictable factors, the STP is important
for both theoretical and practical significance. i
Stochastic programming deals with situ- f (x ) = ;
2
ations where some or all of the parameters of i2 + (x i )

the optimization problem are described by x and i 0and i 0 (1)
random variables rather than by deterministic
quantity. The random variables are defined
as sources, destinations and conveyances An extension of the traditional TP to the
which are depending on the nature and the STP was stated by Shell (Shell, 1955; Haley,
type of problem. Decision making problems 1962) introduced the solution procedure of STP,
of stochastic optimization arise when certain which is an extension of the modified distribu-
coefficients of the optimization model are not tion method. For finding an optimal solution,
fixed or known. In such cases, the quantities the STP requires (m+n+s2) non-zero values
are random. In recent years, methods of multi- of the decision variables to start with a basic
objective stochastic optimization have become feasible solution. (Bit, 1993) developed the
increasingly important in the field of economics, fuzzy programming model for multi-objective
industry, transportation, military purpose and STP. Patel and Tripathy (1989) presented a
technology. The Cauchys distribution is a good computationally superior method for a STP
example of a continuous stable distribution and with mixed constraints (Jimenez, & Verdegay,
this distribution is a source of counter examples, 1998). obtained a solution procedure for uncer-
having little connection with statistical practice. tain STP. (Hitchcock, 1941) first considered the
The Cauchys distribution is an old and very problem of minimizing the cost of distribution
important problem in the statistical literature. of products from several factories to a number
of customers. He developed a procedure to solve Roy (2014) solved multi-choice multi-objective

the TP, which has close resemblance with the transportation problem using utility function ap-
simplex method for solving TP as the primal proach. Tao and Xu (2012) investigated a class
simplex transportation method developed by of rough multiple objective programming and
Dantzig (1963) and Goicoechea et al. (1982) its application to solid transportation problem.
presented the deterministic equivalents for some Messaoudi and Reba (2014) solved stochastic
stochastic programming involving normal and portfolio selection problem in a group decision-
other distribution. Rao (1984) described the making context using fuzzy goal programming
chance constrained and two stage programming approach.
methods for solving a stochastic linear pro- In this paper, we have studied the MCSSTP
gramming problem. Arsham (2012) presented with three stochastic constraints keeping in
the foundation of linear programming with a view the real life problems. The stochastic con-
managerial perspective from solving system straints are inequality type and the parameters
of inequalities to software implementation. (supply, demand and capacity) are followed by
Mahapatra et al. (2010) have presented the Cauchys distribution and cost coefficients of
fuzzy programming technique to obtain the objective function which are under multi-choice
compromise solution of multi-objective sto- environment.
chastic transportation problem. They have also
discussed the conversion procedure from sto-
chastic constraints into deterministic constraints MATHEMATICAL MODEL
when the sources and destination parameters
In the typical STP, we consider the situation
involve randomness. Biswal & Acharya (2009)
having m origins i.e, sources Si (i=1,2,,m)
presented the transformation of a multi-choice
and n demands i.e, destinations Dj (j=1,2,,n).
linear programming problem in which the terms
The sources may be production facilities, ware-
of right hand side in the constraints also satisfy
houses, supply points and the destinations are
the multi-choices or goals. In 2013, Biswal &
consumption facilities, warehouses or demand
Acharya presented a case study on multi-choice
points. Let ek (k=1,2,,s) denotes the unit of
multi-objective mathematical programming
product that can be carried by k different modes
model for integrated production planning. A
of transportation called conveyance, such as
method for modeling the multi-choice goal
land transportation by car or train, airlines and
programming problem, using the multiple terms
ocean shipping.
of binary variables with aspiration levels was
presented by Chang (2007). He has proposed The coefficient C ijk
p
of the objective func-
a revised method for multi-choice goal pro- tion could represent the transportation cost or
gramming model which does not involve any delivery time or unfulfilled supply and demand,
multiplicative terms of binary variables (2008). and others, are provided with transporting a
Chang et al (2010) proposed fuzzy multi-choice unit of product from source i to destination j
goal programming for supplier selection. Hiller by means of the k-th conveyance. Most of real
and Lieberman. (1990) and Ravindran et al. world practical problems are modeled with
(1987) have considered a mathematical model in MCSSTP which are measured in different scales
which an appropriate constraint is to be chosen and in the same type of conflict.
using binary variables. Roy et al. (2012) and In this paper, we have considered the
Mahapatra et al. (2013) discussed the multi mathematical model for MCSSTP involving
choice stochastic transportation problem involv- Cauchys distribution in all constraints and
ing extreme value distribution and exponential cost coefficients of objective function are in
distribution in which the multi-choice concept multi-choice framework as follows.
involved only in the cost parameters. Maity and
Model 1: 2. Only bj (j = 1,2,, n) follows Cauchys

distribution;
m n s 3. Only ek (k = 1,2,, s) follows Cauchys
min : Z = C ijk
1 2
,C ijk p
, ,C ijk { } distribution;
i =1 j =1 k =1 4. All of ai (i = 1,2,, m), bj (j = 1,2,, n)
p
x ijk , p = 1, 2,, , P (2) and ek (k = 1,2,, s) follow Cauchys
distributions.
n s
subject to,Pr(x ijk ai ) 1 i Only ai (i = 1, 2, , m) Follows

j =1 k =1
(3) Cauchys Distribution
i = 1, 2, , m
m s It is assumed that ai (i = 1, 2, , m) are inde-

Pr(x ijk bj ) 1 j pendent random variables having Cauchys

i =1 k =1 distribution in which mean and variance do not
j = 1, 2, , n (4) exist and the i-th variable has two known pa-
rameters i and i (where i = location pa-
m n
rameter and i = scale parameter for sources)
Pr(x ijk ek ) 1 k
which are known to us. The probability den-
i =1 j =1
k = 1, 2, , s (5) sity function of Cauchys distribution for ran-
dom variable ai be given by:
x ijk 0, i , j and k (6)
i
f (a i ) = ;
Here z represents the minimum value of 2
i2 + (a i i )
the objective function. It is assumed that
x and i 0and i 0 (7)
p
ai > 0,bj > 0 andek > 0and C ijk
>0
m n s
The stochastic constraints (3) having ai (i =
and ai bj ek
i =1 j =1 k =1
1, 2,,m) random variables restated as below:
n s
for unbalanced solid transportation problem, Pr(x ijk ai ) 1 i
where specified stochastic levels or predeter-
j =1 k =1
mined confidence levels are defined as 0 < i i = 1, 2, , m

< 1, i; 0 < j < 1, j, and 0 < k < 1, k.
We assumed that ai (i = 1, 2, ,m), bj (j Hence the above stochastic constraints can
= 1, 2, , n) and ek (k = 1, 2, , s) are speci- be reduced to the cumulative density function
using the probability density function of Cau-
fied with Cauchys distribution and C p
ijk > 0,
chys distribution as follows:
p are also defined as multi-choice parameters
in each respective decision variable. Now the
following cases are to be considered:
f (ai )d (ai ) 1 i , i = 1, 2, , m (8)
n s
1. Only ai (i = 1,2,, m) follows Cauchys xijk

j =1 k =1
distribution;
m s
i
Pr(x ijk bj ) 1 j
2
n s
i2 + (a i i ) i =1 k =1
x ijk j = 1, 2, , n (13)
j =1 k =1
d (ai ) 1 i , i = 1, 2, , m (9)
Hence the above stochastic constraints can
be reduced to the cumulative density function
On simplification, we have: using probability density function of Cauchys
distribution as follows:
n s

x ijk i i tan i
j =1 k =1 2
(10) f (bj )d (bj ) 1 j ;

i = 1, 2, , m m s

xijk
Finally, the stochastic constraints (3) can i =1 k =1
j = 1, 2, , n
be expressed into equivalent deterministic
constraints as follows:
which can be further simplified to:
n s
x ijk i + i tan i
2 j

j =1 k =1
i = 1, 2, , m (11) 2
m
xijk
s

(
j 2 + b j j

)
Only bj (j = 1, 2,,n) Follows i =1 k =1
d (bi ) 1 j j = 1, 2, , n (14)
Cauchys Distribution
Assume that b j (j = 1, 2, , n) are indepen-

On simplification, we have:
dent random variables having Cauchys
distribution in which mean and variance do
m s
not exist and the j-th random variable has tan
ijk x
j j 2
j
two known parameters j' and j (where i =1 k =1
j' = location parameter and j = scale j = 1, 2, , n (15)
parameter for destinations) which are known
Finally, the stochastic constraints (4) can
to us. The probability density function of
be expressed into equivalent deterministic
Cauchys distribution for random variable
constraints as follows:
bj be given by:
m s
j
x ' j + ' j tan j
f (b j ) = ; ijk
2
2 i =1 k =1
j 2 + (bj j ) j = 1, 2, , n (16)

x and j 0and
j 0 (12)
Only ek (k = 1,2,, s) Follows
Cauchys Distribution
The stochastic constraints (4) having bj (j = Assume that ek (k = 1, 2, , s) are independent
1,2,, n) random variables restated as follows: random variables having Cauchys distribution
in which mean and variance do not exist and Finally, the stochastic constraints (5) can
the k-th random variable has two known pa- be expressed into deterministic constraints as
rameters k'' and k (where k'' = location pa- follows:
rameter and k (= scale parameter for capaci-
ties) which are known to us. The probability
m n
x k'' +k'' tan k
density function of Cauchys distribution for i =1 j =1
ijk
2
random variable ek be given by: k = 1, 2, , s (21)
k All of ai (i = 1, 2,,m), bj
f (e k ) = ;
2
(j = 1, 2,,n) and ek (k = 1, 2,,s)
k 2 + ek k

( ) Follow Cauchys Distributions
x and k 0and
k 0 (17)
In this case, the equivalent deterministic
model of MCSSTP with specified con-
The stochastic constraints (5) having ek (k fidence levels i , j and k can be
= 1,2,, s) random variables restated as below: represented as Model 2 in the following
section.
m n
Pr(x ijk ek ) 1 k Model 2:

i =1 j =1
k = 1, 2, , s
m n s
min :Z = C ijk
1 2
,C ijk p
{
, ,C ijk }
Hence the above stochastic constraints i =1 j =1 k =1
can be reduced to cumulative density function

p
x ijk , p = 1, 2, , P (22)
using probability density function for Cauchys
distribution as:
n s
x i + i tan i
ijk
2
f (ek )d (ek ) 1 k ;
j =1 k =1
i = 1, 2, , m (23)
m n

x ijk
i =1 j =1
k = 1, 2, , s (18) n s
x ' j + ' j tan j
j =1 k =1
ijk

2
j = 1, 2, , n (24)
k
d (ek ) k ;

( )
2
k 2 + ek k m n
k'' +k'' tan k
m n
xijk
i =1 j =1
x ijk
2
k = 1, 2, , s (19) i =1 j =1
k = 1, 2, , s (25)
On simplification, we have:
x ijk 0, i , j and k (26)
m n
x k'' k'' tan k
i =1 j =1
ijk
2
k = 1, 2, , s (20)
TRANSFORMATION OF AN m n s
EQUIVALENT MODEL WITH min :Z = C ijk

1 2
,C ijk {
C 3 x ijk
, ijk }
COST COEFFICIENTS OF i =1 j =1 k =1
OBJECTIVE FUNCTION
The cost coefficients have three
The proposed model is specified for twenty
choices on the cost coefficients of objective choices namely, C ijk
1 2
,C ijk{ C 3 out of which
, ijk }
function. For p = 1, the case is trivial. Nineteen one is to be selected. Since 21 < 3 < 22 ,
cases are expressed in the following form for so the total number of elements of the set
p = 2, 3, , 20: are 3. Denoting the binary variables as:
1
Z ijk 2
,Z ijk and introducing auxiliary con-
Step 1: When p = 2 straint in two models are formulated as
below.
We rewrite the objective function (2) as
follows: Model 2(a):
m n s
{ }
m n s
min :Z = C ijk
1 2
,C ijk min : Z =
i =1 j =1 k =1
i =1 j =1 k =1
x ijk , p = 1, 2, , P
ijk (
C 1 1 Z 1 1 Z 2
ijk )( ijk )
1
+
ijk

Z (
1 Z 2
ijk )
C 2
ijk x ijk

The cost coefficients have two choices
namely, {C ijk 1 2
,C ijk }, out of which one is to be ijk (
+Z 2 1 Z 1 C 3
ijk ) ijk

selected. Since total number of elements of the 1
Z ijk 2
+ Z ijk 1
set are 2, so only one binary variable is required.
Denoting the binary variable Z ijk 1
, the above x ijk 0, i , j and k
equation is formulated as below: q
Z ijk = 0 / 1,q = 1, 2
m n s
min :Z = Model 2(b):

i =1 j =1 k =1
{ 1
C ijk 1
Z ijk (
+ 1Z 1
ijk )C } x
2
ijk ijk
m n s
min : Z =
x ijk 0, i , j and k i =1 j =1 k =1
C 1 Z 1 Z 2
ijk ijk ijk
+Z 1 1 Z 2 C 2 x
and: ijk
( ijk )
ijk
ijk
+Z 2 1 Z 1 C 3
( )
ijk ijk ijk
q
Z ijk = 0 / 1, q = 1 1
Z ijk 2
+ Z ijk 1
Step 2: When p = 3 x ijk 0, i , j and k

q
Z ijk = 0 / 1,q = 1, 2
follows:
Step 3: When p = 4 The cost coefficients have five choices as

{ 12
C ,C ijk
ijk
C 3 ,C ijk
, ijk
4 5
,C ijk out of which one is to }
We rewrite the objective function (2) as be selected. Since 2 <5 <23, so we need three2
follows:
binary variables and they are Z ijk1 2
,Z ijk 3
,Z ijk .
m n s Then we put restriction to remaining three terms
min :Z = C ijk
1 2
,C ijk C 3 ,C ijk
, ijk
4
{
x ijk } (8 5) by introducing auxiliary and additional
i =1 j =1 k =1 constraints in three different models which are
expressed as given below.
The cost coefficients of the objective func-
t i o n h a v e f o u r c h o i c e s n a m e l y, Model 4(a):
{ 1
C ijk 2
,C ijk C 3 ,C ijk
, ijk
4
}
out of which one is to be m n s
selected. Since the total number of choices are min :Z =
4 = 22. Denoting the binary variables as: Z ijk
1 2
,Z ijk i =1 j =1 k =1
so we construct the objective function in the

1
{Z ijk 2
(1 Z ijk 3
) 1 Z ijk ( )
following form. C 1
ijk
+Z (1 Z )(1 Z )C
2
ijk
1
ijk
3
ijk
2
ijk
Model 3:
3
+Z ijk (1 Z )(1 Z )C
1
ijk
2
ijk
3
ijk
1
+Z ijk Z (1 Z )C
2
ijk
3
ijk
4
ijk
m n s
min : Z = 2
+Z ijk Z (1 Z )C }x
3
ijk
1
ijk
4
ijk ijk
i =1 j =1 k =1
Z Z C
1 2 1
1 Z 1
+Z 2
+Z 3
1
ijk ijk ijk

ijk ijk ijk

(
+Z ijk 1 Z C 2
)
ijk
2
ijk

x
2 Z 1
ijk
+Z 3
ijk
1
+Z ijk

2
( 1
1 Z ijk )3
C ijk ijk
x ijk 0, i , j and k
4
( + 1 Z ijk

1
)( 2
1 Z ijk C ijk )

q
Z ijk = 0 / 1,q = 1, 2, 3
q
Z ijk = 0 / 1,q = 1, 2 Model 4(b):
m n s
Step 4: When p = 5 min :Z =

i =1 j =1 k =1

1
{Z ijk 2
(1 Z ijk 3
) 1 Z ijk ( )
follows: C 1
ijk
+Z (1 Z )(1 Z )C
2
ijk
1
ijk
3
ijk
2
ijk
m n s
3
+Z ijk (1 Z )(1 Z )C
1
ijk
2
ijk
3
ijk
min :Z =

1
+Z ijk Z (1 Z )C
2
ijk
3
ijk
4
ijk
i =1 j =1 k =1
{C 1
ijk
2
,C ijk 3
,C ,C ,C
ijk
4
ijk
5
ijk }x ijk 2
+Z ijk Z (1 Z )C }x
3
ijk
1
ijk
4
ijk ijk
1 2 3 Then we put restriction to remaining two terms

1 Z ijk + Z ijk + Z ijk
3 2
(8 6) by introducing auxiliary and additional
2 Z ijk + Z ijk 1 constraints in three different models which are
expressed as given below.
q
Z ijk = 0 / 1,q = 1, 2, 3 Model 5:
m n s
Model 4(c): min :Z =
i =1 j =1 k =1
m n s
1
{Z ijk 2
(1 Z ijk 3
) 1 Z ijk ( )
min :Z =
i =1 j =1 k =1
1
C ijk ( 2
+ Z ijk )( 1
1 Z ijk ) 3
1 Z ijk 2
C ijk
{Z 1
ijk
(1 Z ) 1 Z 2
ijk ( 3
ijk )
+ (1 Z ) (1 Z ) Z C
1 2 3 3
(1 Z )(1 Z )C
ijk ijk ijk ijk
C 1
+Z 2 1 3 2
ijk ijk ijk ijk ijk +Z Z (1 Z )C
1
ijk
2
ijk
3
ijk
4
ijk
+Z (1 Z ) (1 Z )C
3
ijk
1
ijk
2
ijk
3
ijk +Z Z (1 Z )C
1 3 2 5
ijk ijk ijk ijk
+ (1 Z ) Z Z C 1
2 3 4
ijk ijk ijk ijk +Z Z (1 Z )C }x
2
ijk
3
ijk
1
ijk
6
ijk ijk
+Z (1 Z ) Z C
1
ijk
}x 2
ijk
3
ijk
5
ijk ijk 1
1 Z ijk 2
+ Z ijk 3
+ Z ijk 2
1 2 3
1 Z +Z +Z
ijk
1
ijk
2
ijk
2Z ijk
+Z ijk
1
q
Z ijk = 0 / 1,q = 1, 2, 3
q
Z ijk = 0 / 1,q = 1, 2, 3 Step 6: When p = 7
Step 5: When p = 6 We rewrite the objective function (2) as

follows:
m n s
follows:
min :Z =
i =1 j =1 k =1
min :Z =
m n s
{C 1
ijk
2 3
,C ,C ,C ,C ,C ,C
ijk ijk
4
ijk
5
ijk
6
ijk
7
ijk }x ijk
i =1 j =1 k =1
{C 1
ijk
2
,C ijk 36
,C ,C ,C ,C ijk
ijk
4
x ijk
ijk
5
ijk } The cost coefficients have seven
choices as C ijk
1 2
,C ijk {
C 3 ,C ijk
, ijk
4 5
,C ijk 6
,C ijk 7
,C ijk }
The cost coefficients have six choices as out of which one is to be selected. Since
{ 1 2
C ijk ,C ijk C 3 ,C ijk
, ijk
4 5
,C ijk 6
,C ijk out of which one } 2 2 <7 <2 3 , so we need three binary vari-
is to be selected. Since 22 <6 <23, so we need ables and they are Z ijk
1 2
,Z ijk 3
,Z ijk . Then we
three binary variables and they are Z ijk
1 2
,Z ijk 3
,Z ijk . put restriction to remaining two terms
(8 7) by introducing auxiliary and ad- q

Z ijk = 0 / 1,q = 1, 2, 3
ditional constraints in three different
models which are expressed as given
below. Step 7: When p=8
Model 6(a): We present the objective function (2) as

follows:
m n s
min :Z = m n s
i =1 j =1 k =1 min :Z =
1
{(1 Z ijk 3
)(1 Z ) 1 Z ijk 1
C ijk 2
( ) i =1 j =1 k =1
C 1 ,C 2 ,C 3 ,C 4 ,
ijk
1
+Z (1 Z ) 1 Z
ijk )C 2
ijk ( 3
ijk
2
ijk
ijk ijk ijk ijk
5 8 ijk
x
C ijk 6
,C ijk 7
,C ijk ,C ijk
+ (1 Z )C + (1 Z ) (1 Z )
3
ijk
3
ijk
1
ijk
2
ijk

Z C + Z Z (1 Z )C
3 4 1 2 3 5
ijk ijk ijk ijk ijk ijk
The cost coefficients have eight choices as:
+Z Z (1 Z )C
1
ijk
3
ijk
2
ijk
6
ijk

+Z Z (1 Z )C }x
2
ijk
3
ijk
1
ijk
7
ijk ijk
{C 1
ijk
2
,C ijk C 3 ,C ijk
, ijk
4 5
,C ijk 6
,C ijk 7
,C ijk C8
, ijk }
1
Z ijk 2
+ Z ijk 3
+ Z ijk 2
out of which one is to be selected. Since 8 = 23,
x ijk 0, i , j and k so we need three binary variables and they are
1
Z ijk 2
,Z ijk 3
,Z ijk . Therefore we have formulated
q
Z ijk = 0 / 1,q = 1, 2, 3
only one model as given below:
Model 6(b): m n s
min :Z =
m n s i =1 j =1 k =1
min :Z =
1
{(1 Z ijk 2
)(1 Z ijk 8
)C ijk
i =1 j =1 k =1
+Z 1
( )C
(1 Z ) 1 Z 2 3 7
1
{Z ijk (
(1 Z ) 1 Z )C 2
ijk
3
ijk
1
ijk

ijk ij k ijk ijk
+Z (1 Z ) (1 Z )C
2 1 3 2 + (1 Z ) Z (1 Z )C
1
ijk
2
ijk
3
ijk
6
ijk
ijk ijk ijk ijk
+ (1 Z ) (1 Z ) Z C
1 2 3 5
+ (1 Z ) (1 Z ) Z C 1 2 3 3 ijk ijk ijk ijk

ijk ijk ijk ijk
+Z Z (1 Z )C
+Z Z (1 Z )C
1 2 3 4
1 2 3 4
ijk ijk ijk ijk
ijk ijk ijk ijk
+Z Z (1 Z )C
1 3 2 3
+Z Z (1 Z )C
1 3 2 5 ijk ijk ijk ijk
ijk ijk
ijk ijk
+Z Z (1 Z )C
2 3 1 2
+Z Z (1 Z )C
2 3 1 6 ijk ijk ijk ijk
ijk ijk ijk ijk
1
+Z ijk 2
Z ijk 3
Z ijk 1
C ijk }x ijk
1
+Z ijk 2
Z ijk 3
Z ijk 7
C ijk }x ijk
1
Z ijk 2
+ Z ijk 3
+ Z ijk 1
q
Z ijk = 0 / 1,q = 1, 2, 3
Step 8: When p=9 1

+(1 Z ijk 2
)Z ijk 3
1 Z ijk (4
Z ijk 6
C ijk )
We present the objective function (2) as +(1 Z )Z Z 1
ijk
2
ijk
3
ijk (1 Z )C 4
ijjk
7
ijk
follows: 1 2 3 4 8
+Z ijk (1 Z ijk )(1 Z ijk )Z ijkC ijk

m n s
1
+ Z ijk 2
(1 Z ijk 3
)Z ijk 4
1 Z ijk 9
C ijk }x ijk ( )
min :Z =
i =1 j =1 k =1
1
1 Z ijk 2
+ Z ijk 3
+ Z ijk 4
+ Z ijk 2

C ,C ,C ,C ,
1 2
3 4

ijk ijk
x
ijk ijk 1
Z ijk 2
+ Z ijk 1
C ,C ,C ,C ,C ijk
5 6 7 8 9
ijk ijk ijk ijk ijk
The cost coefficients of the objective

q
Z ijk = 0 / 1,q = 1, 2, 3, 4
function have nine choices as:
Model 8(b):
{C 1
ijk
2
,C ijk C 3 ,C ijk
, ijk
4 5
,C ijk 6
,C ijk 7
,C ijk C8 C9
, ijk , ijk }
m n s
min :Z =
out of which one is to be selected. Since i =1 j =1 k =1
2 3 < 9 <2 4, so we need four binary variables 1
{(1 Z ijk 2
) 1 Z ijk( 3
Z ijk 4
Z ijk 1
C ijk )
and they are Z ijk 1 2
,Z ijk 3
,Z ijk , Z ijk
4
. Then we
put the restriction to the introducing aux-
(
+ 1Z 1
)Z
(1 Z )Z C
ijk
2
ijk
3
ijk
4
ijk
2
ijk
iliary and additional constraints in the + (1 Z Z (1 Z )C

1
ijk )Z 2
ijk
3
ijk
4
ijk
3
ijk
mathematical model. Using the four bi-
nary variables with additional restrictions +Z (1 Z ) (1 Z ) Z C
1
ijk
2
ijk
3
ijk
4
ijk
4
ijk
we formulate two different models. Again

+Z (1 Z )Z (1 Z )C
1 2 3 4 5
for each model, there are six similar mod- ijk ijk ijk ijk ijk

els. To reduce the length of the paper, we +Z Z (1 Z ) (1 Z )C
1
ijk
2
ijk
3
ijk
4
ijk
6
ijk
have stated only two models as given 1 2 3 4 7
below. +(1 Z ijk )Z ijk Z ijk Z ijkC ijk
1 2 3 4

+ Z ijk (1 Z ijk )Z ijk Z ijkC i8jk
Model 8(a): 1
+Z ijk 2
Z ijk ( 3
1 Z ijk 4
Z ijk 9
C ijk )
}x ijk
m n s
min :Z =
1
2 Z ijk 2
+ Z ijk 3
+ Z ijk 4
+ Z ijk 3
i =1 j =1 k =1 1
Z ijk 2
+ Z ijk 3
+ Z ijk 2
1
{(1 Z ijk 2
)(1 Z ijk 3
) 1 Z ijk (4
Z ijkC ij1k )
1
+(1 Z ijk 2
) (1 Z ijk 3
)Z ijk x ijk 0, i , j and k
(1 Z )C + (1 Z ) Z = 0 / 1,q = 1, 2, 3, 4
q
4 2 1 2 Z ijk
ijk ijk ijk ijk
(1 Z )(1 Z )C
3
ijk
4
ijk
3
ijk

+Z (1 Z ) (1 Z ) (1 Z )C
1
ijk
2
ijk
3
ijk
4
ijk
4
ijk
Step 9: When p = 10
1 2 4 5 3
+(1 Z ijk )(1 Z ijk )Z Z
Z ijkC ijk ijk
We rewrite the objective function (2) as follows:
m n s Model 9(b):
min :Z =
i =1 j =1 k =1 m n s
{ 1
C ,C , ,C
ijk
2
ijk
10
ijk } ijk
x min :Z =
i =1 j =1 k =1
1
{(1 Z ijk 2
) 1 Z ijk ( 3
Z ijk 4
Z ijk 1
C ijk )
(
+ 1Z 1
(1 Z )Z C)Z 2 3 4 2
tion have ten choices as C ijk

1 2
,C ijk ,
10
,C ijk out { } ijk ijk ijk ijk ijk
of which one is to be selected. Since 23< 10 + (1 Z Z (1 Z )C

1
ijk )Z 2
ijk
3
ijk
4
ijk
3
ijk

<24, so we need four binary variables and they +Z (1 Z ) (1 Z ) Z C
1
ijk
2
ijk
3
ijk
4
ijk
4
ijk
are Z ijk
1 2
,Z ijk 3
,Z ijk , Z ijk
4
. We express the two sets
of binomial coefficients as {4C1, 4C2} and {4C1, +Z (1 Z )Z (1 Z )C
1
ijk
2
ijk
3
ijk
4
ijk
5
ijk

4
C3} whose individual sum is equal to 10 for +Z Z (1 Z ) (1 Z )C
1
ijk
2
ijk
3
ijk
4
ijk
6
ijk
the same goal as p = 10. So there does not arise
1 2 3 4 7
any additional constraint. Two different models +(1 Z ijk )Z ijk Z ijk Z ijkC ijk

are formulated as given below: 1
+ Z ijk 2
(1 Z ijk 3
)Z ijk 4
Z ijkC ij8jk
Model 9(a):
1
+Z ijk 2
Z ijk ( 3
1 Z ijk 4
Z ijk 9
C ijk )
m n s
1
+Z Z Z
ijk
2
ijk
3
ijk (1 Z )C 4
ijk
10
ijk
}x ijk
min :Z = 1
2 Z ijk 2
+ Z ijk 3
+ Z ijk 4
+ Z ijk 3
i =1 j =1 k =1
1
{(1 Z ijk 2
)(1 Z ijk 3
) 1 Z ijk 4
Z ijkC ij1k ( ) x ijk 0, i , j and k
1 2 3
+(1 Z ijk ) (1 Z ijk )Z ijk q
Z ijk = 0 / 1,q = 1, 2, 3, 4
(1 Z )C + (1 Z )Z
4
ijk
2
ijk
1
ijk
2
ijk
(1 Z )(1 Z )C
3
ijk
4
ijk
3
ijk
Step 10: When p = 11
+Z (1 Z ) (1 Z ) (1 Z )C
1 2 3 4 4
ijk ijk ijk ijk ijk

1
+(1 Z )(1 Z )Z Z C 2 3 4 5 follows:
ijk ijk ijk ijk ijk
1
+(1 Z ijk 2
)Z ijk 3
1 Z ijk (
4
Z ijk 6
C ijk )
m n s
min : Z =
1
+(1 Z ijk 2
)Z ijk 3
Z ijk 1 Z ij4jk C ijk
7
( ) i =1 j =1 k =1
1
+Z ijk 2
(1 Z ijk 3
)(1 Z ijk 4
)Z ijk 8
C ijk {C 1
ijk
2
,C , .,C
ijk
11
ijk ijk }x

1
+Z ijk 2
(1 Z ijk 3
)Z ijk 4
1 Z ijk 9
C ijk ( )
1
+Z ijk 2
Z ijk 3
(
1 Z ijk )(1 Z )C 4
ijk
10
ijk
}x ijk tion have eleven choices as C ijk
1 2
,C ijk ,
11
, ,C ijk { }
1 Z 1
+Z 2
+Z 3
+Z 4
2 out of which one is to be selected. Since 23< 11
ijk ijk ijk ijk
<24, so we need four binary variables and they
x ijk 0, i , j and k are Z ijk
1 2
,Z ijk 3
,Z ijk , Z ijk
4
. We express the two set
q
Z ijk = 0 / 1,q = 1, 2, 3, 4 of binomial coefficients as {4C0,4C1,4 C2 } and
{4C1,4C2,4 C3} in respect to two possible models.
In each model, the auxiliary constraints depend

upon the above specified sets of binomial coef- ( 1
+ 1 Z ijk 2
Z ijk 3
Z ijk ) 4
1 Z ijk 3
C ijk ( )
ficients and its ranges. The sum of terms of +Z 1
ijk (1 Z )(1 Z ) Z C
2
ijk
3
ijk
4
ijk
4
ijk
binomial coefficients is just equal to 11, which
is the same of the aspiration level as p = 11. So
1
+Z ijk (1 Z )Z (1 Z )C
2
ijk
3
ijk
4
ijk
5
ijk

there does not arise any additional constraint. 1
+Z ijk Z (1 Z ) (1 Z )C
2
ijk
3
ijk
4
ijk
6
ijk
Two different models are formulated as given
1 2 3 4 7
below. +(1 Z ijk )Z ijk Z ijk Z ijkC ijk
1 2 3 4

+ Z ijk (1 Z ijk )Z ijk Z ijkC ij8jk
Model 10(a):
1
+Z ijk 2
Z ijk (
3
1 Z ijk 4
Z ijk 9
C ijk )
min :Z =
m n s
+Z Z Z 1
ijk
2
ijk
3
ijk (1 Z )C 4
ijk
10
ij k

i =1 j =1 k =1 1
+Z ijk 2
Z ijk 3
Z ijk 4
Z ijk 11
C ijk }x ijk
1
{(1 Z ijk 2
)(1 Z ijk 3
) 1 Z ijk 4
1 Z ijk ( 1
C ijk )( ) 1
Z ijk 2
+ Z ijk 3
+ Z ijk 4
+ Z ijk 2
+(1 Z )(1 Z 1 2
) (1 Z ) 3
4 2
ijk
1
ijk
ijk
2
Z ijkC ijk + (1 Z ijk ) (1 Z ) ijk
q
Z ijk = 0 / 1,q = 1, 2, 3, 4
3
Z ijk ( 4
1 Z ijk 3
C ijk ) 1
+ (1 Z ijk 2
)Z ijk

(1 Z )(1 Z )C + Z
3
ijk
4
ijk
4
ijk
1
ijk
(1 Z )(1 Z )(1 Z )C
2
ijk
3
ijk
4
ijk
5
ijk

1 2 3 4 6 We rewrite the objective function (2) as
+(1 Z ijk )(1 Z ijk )Z ijk Z C ijk ijk
follows:
1
+(1 Z ijk 2
)Z ijk 3
1 Z ijk ( 4
Z ijk 7
C ijk )
(1 Z )C
m n s
1 2 3 4 8
+(1 Z )Z Z ijk ijk ijk ijjk ijk min :Z =
i =1 j =1 k =1
1 2 3 4 9
{C 1 2
,C , .,C 12
}x
( )
ijk ijk ijk ijk
1 2 3 4 10
+Z ijk (1 Z ijk )Z ijk 1 Z ijk C ijk
1
+Z ijk 2
Z ijk ( 3
1 Z ijk )(1 Z )C 4
ijk
11
ijk
}x ijk The cost coefficients of the objective
function have twelve choices as
2
{ }
1 2 3 4
Z ijk + Z ijk + Z ijk + Z ijk 1
C ijk 2
,C ijk 12
, ,C ijk out of which one is to be
,
x ijk 0, i , j and k selected. Since 23< 12<24, so we need four

binary variables and they are Z ijk
1 2
,Z ijk 3
,Z ijk ,
q
Z ijk = 0 / 1,q = 1, 2, 3, 4 4
Z ijk . We consider the set of binomial coef-
ficients as {4C1, 4C2, 4 C3 } and the sum of
Model 10(b): terms is 14, which is just greater than the
selected aspiration level or goal as p = 12.
m n s The auxiliary constraints depend upon the
min :Z = above specified set of binomial coefficient
i =1 j =1 k =1 and its ranges, and two (1412 = 2) addi-
1
{(1 Z ijk 2
) 1 Z ijk( 3
Z ijk 4
Z ijk 1
C ijk ) tional constraints are performed. Total six
(
+ 1Z 1
ijk )Z 2
ijk
(1 Z )Z C 3
ijk
4
ijk
2
ijk
similar models are possible for this aspiration
level since p = 12. To reduce the length of m n s

the paper, we have formulated only one min :Z =
model as given below. i =1 j =1 k =1
{C 1
ijk
2
,C , .,C
ijk
13
ijk }x
ijk
Model 11:
m n s
min :Z = tion have thirteen choices as C ijk
1 2
,C ijk ,
13
, ,C ijk { }
i =1 j =1 k =1
out of which one is to be selected. Since 23<
1
{(1 Z ijk 3
)(1 Z ) 1 Z ijk 4
Z ijkC ij1k
2
ijk ( ) 13<24, so we need four binary variables and
1 3 2
+(1 Z ijk ) (1 Z )Z ijk ijk they are Z ijk
1 2
,Z ijk 3
,Z ijk , Z ijk
4
. We consider the
set of consecutive terms of binomial coefficients
(1 Z )C + (1 Z )Z
4
ijk
2
ijk
1
ijk
2
ijk
{4C1, 4C2, 4 C3 } and their sum is exactly equal
(1 Z )(1 Z )C + Z
3
ijk
4
ijk
3
ijk
1
ijk to 14 which is just greater than the selected
aspiration level or goal as ki= 13. Therefore,
(1 Z )(1 Z )(1 Z )C
2
ijk
3
ijk
4
ijk
4
ijk
exactly one auxiliary constraint can be formu-
+Z Z (1 Z ) (1 Z )C
1
ijk
2
ijk
3
ijk
4
ijk
5
ijk
lated with the help of above binomial coeffi-
cients and its ranges. Then we put the restriction
+Z (1 Z )Z (1 Z )C
1
ijk
2
ijk
3
ijk
4
ijk
6
ijk
to remaining one (i.e, 14 13 = 1) term by
introducing additional constraint in the model.
+(1 Z )Z (1 Z ) Z C2 2 3 4 7
ijk ijk ijk ijk ijk
Here four similar models are possible for this
1
+(1 Z ijk 2
)(1 Z ijk 3
)Z ijk 4
Z ijk 8
C ijk aspiration level since p = 13. Here, we have
formulated only one model as given below:
2
+Z ijk 1
(1 Z ijk 3
)Z ijk 1 Z ij4jk C ijk
9
( )
( )(1 Z )C
3 2 1 4 10 m n s
+Z ijk Z ijk 1 Z ijk min :Z =
1 2 3
ijk ijk

+Z ijk (1 Z ijk )Z ijk Z ij4kC ijk
11
i =1 j =1 k =1
1
+Z ijk 3
(1 Z ijk 2
)Z ijk 4
Z ijk 12
C ijk }x ijk
1
{Z ijk 1
4
(1 Z ijk 2
)(1 Z ijk 3
) 1 Z ijk C i1jk ( )
2 1
+Z ijk (1 Z ijk )
1
1 Z ijk 2
+ Z ijk 3
+ Z ijk 4
+ Z ijk 3
(1 Z )(1 Z )C
3
ijk
4
ijk
2
ijk

2 +(1 Z )Z (1 Z ) (1 Z )C
2 3 4
Z ijk
+Z ijk
+Z ijk
1 3 2 4 3
ijk ijk ijk ijk ij k
1
Z ijk 2
+ Z ijk 3
+ Z ijk 2
+(1 Z ) (1 Z ) (1 Z ) Z C
1 3 2 4 4
ijk ijk ijk ijk ijk

x ijk 0, i , j and k 1
+ Z ijk 2
Z ijk 3
(1 Z ijk 4
)(1 Z ijk 5
)C ijk
q
Z ijk = 0 / 1,q = 1, 2, 3, 4 1
+Z ijk 3
Z ijk 2
(1 Z ijk 4
) 1 Z ijk 4
(
Z ijk 6
C ijk )
1
+Z ijk 4
Z ijk 2
(1 Z ijk 3
) 1 Z ijk 7
C ijk( )
Step 12: When p = 13 2 3 1 4 8
+Z ijk Z ijk (1 Z ijk )(1 Z ijk )C ijk

1
+(1 Z ijk 3
) 1 Z ijk (
Z ij2k Z ijk
4 9
C ijk )
follows:
3
+Z ijk 4
Z ijk 1
1 Z ijk( 2
1 Z ijk 10
C ijk )( )
(1 Z )(1 Z )C
3
ijk
4
ijk
2
ijk

1
+Z ijk 4
(1 Z ijk 3
)Z ijk Z ij2kC ijk
11
+(1 Z )Z (1 Z ) (1 Z )C
1
ijk
3
ijk
2
ijk
4
ijk
3
ij k
+(1 Z ) (1 Z ) (1 Z ) Z C
1 3 2 4 12
+Z ijk (1 Z ijk )Z ijk Z ijkC ijk 1 3 2 4 4
ijk ijk ijk ijk ijk

+Z Z Z 1
ijk
4
ijk
3
ijk (1 Z )C2
ijk
13
ij k
}x ijk 1
+ Z ijk 2
Z ijk 3
(1 Z ijk 4
)(1 Z ijk 5
)C ijk
1
1 Z ijk 2
+ Z ijk 3
+ Z ijk 4
+ Z ijk 3 1
+Z ijk 3
Z ijk 2
(1 Z ijk 4
) 1 Z ijk 4
(
Z ijk 6
C ijk )
Z 2
ijk
+Z 3
ijk
+Z 4
ijk
2
1
+Z ijk 4
Z ijk 2
(1 Z ijk 3
) 1 Z ijk 7
C ijk( )
2 3 1 4 8
x ijk 0, i , j and k +Z ijk Z ijk (1 Z ijk )(1 Z ijk )C ijk

Z q
= 0 / 1,q = 1, 2, 3, 4
1
+(1 Z ijk 3
) 1 Z ijk (
Z ij2k Z ijk
4 9
C ijk)
ijk
3
+Z ijk 4
Z ijk 1
(
1 Z ijk )(1 Z )C 2
ijk
10
ijk

Step 13: When p = 14 1
+Z ijk 4
(1 Z ijk 3
)Z ijk Z ij2kC ijk
11
1 3 2 4 12
We rewrite the objective function (2) as +Z ijk (1 Z ijk )Z ijk Z ijkC ijk

follows: 1
+Z ijk 4
Z ijk 3
Z ijk 2
1 Z ijk(C i13jk )
m n s
1
+(1 Z ijk 2
)Z ijk 3
Z ijk 4
Z ijk 14
C ijk }x ijk
min :Z =
i =1 j =1 k =1
1
1 Z ijk 2
+ Z ijk 3
+ Z ijk 4
+ Z ijk 3
{C 1
ijk
2
,C ijk , .,C 14
ijk}x ijk x ijk 0, i , j and k
q
Z ijk = 0 / 1,q = 1, 2, 3, 4
tion have fourteen choices as C ijk
1 2
,C ijk ,
14
, ,C ijk { }
14<24, so we need four binary variables and
they are Z ijk
1 2
,Z ijk 3
,Z ijk , Z ijk
4
. We recall the rela-
follows:
tion 24 =4C0 +4C1+4C2+ 4 C3+ 4 C4, and construct
a set whose elements are the consecutive bino- m n s
mial coefficients i.e, {4C1, 4C2, 4 C3}, and the min :Z =
sum is 14 which is equal to the goal since p = i =1 j =1 k =1
14. So there does not arise any additional con-
straint. Hence only one model is formulated as
{C 1
ijk
2
,C , .,C
ijk
15
ijk }x
ijk
given below:
min :Z =
m n s
tion have fifteen choices as C ijk
1 2
,C ijk ,
15
, ,C ijk { }
i =1 j =1 k =1
1
{Z ijk 1
4
(1 Z ijk 2
)(1 Z ijk 3
) 1 Z ijk C i1jk ( ) 15<24, so we need four binary variables and
2 1 they are Z ijk
1 2
,Z ijk 3
,Z ijk , Z ijk
4
. We express 15 as
+Z ijk (1 Z ijk )
two sets of consecutive terms of binomial coef- Model 14(b):

ficients {4C0, 4C1, 4C2, 4 C3 } or {4C1, 4C2, 4 C3,
4
C4 } in respect to two different models. The m n s
auxiliary constraints are performed with the min :Z =

help of above two sets of binomial coefficients i =1 j =1 k =1
with their restriction as sum of four binary

1
{(1 Z ijk 2
)(1 Z ijk 3
) 1 Z ijk 4
Z ijkC ij1k ( )
variables does not equal to 0 and product of 1
+Z (1 Z )Z 2 3
ijk ijk ijk
them is not equal to 1. Therefore two differ-
ent models are formulated as given below. (1 Z )C + (1 Z )
4
ijk
2
ijk
1
ijk
Z (1 Z ) (1 Z )C
2
ijk
3
ijk
4
ijk
3
ijk
Model 14(a):
+(1 Z )Z (1 Z ) (1 Z )C
1 2 3 4 4
m n s 2
ijk
1 4
ijk ijk
3 5
ijk ijk

+(1 Z )(1 Z ijk )Z Z ijkC
min :Z = ijk ijk ijk
1
{(1 Z ijk 2
)(1 Z ijk
i =1 j =1 k =1
3
) 1 Z ijk 4
Z ijkC ij1k ( )
1
+(1 Z ijk 2
)Z ijk 3
1 Z ijk 4
Z ijk 6
C ijk( )
1
+Z ijk 2
(1 Z ijk 3
)Z ijk
+(1 Z )Z Z 1
ijk
2
ijk
3
ijk (1 Z )C 4
ijjk
7
ijk
1 2 3 4 8
(1 Z )C + (1 Z )
4 2 1

( )
ijk ijk ijk
1 2 3
+Z ijk (1 Z ijk )Z ijk 1 Z i4jk C ijk
9
Z (1 Z ) (1 Z )C
2
ijk
3
ijk
4
ijk
3
ijk
+(1 Z )Z (1 Z ) (1 Z )C
1 2 3 4 4
1
+Z ijk 2
Z ijk 3
1 Z ijk( )(1 Z )C 4
ijk
10
ijk

ijk
1
ijk
2
ijk
3 4 5
ijk ijk
1
+Z ijk 2
(1 Z ijk 3
)Z ijk Z ij4kC ijk
11
+(1 Z )(1 Z )Z Z C
ijk ijk ijk ijk ijk
1 3 2 4 12
+Z ijk (1 Z ijk )Z ijk Z ijkC ijk
1
+(1 Z ijk 2
)Z ijk 3
1 Z ijk 4
Z ijk (6
C ijk )
+(1 Z )Z Z 1
ijk
2
ijk
3
ijk (1 Z )C 4
ijjk
7
ijk
1
+Z ijk 2
Z ijk 3
Z ijk 4
1 Z ijk (
C i13jk )
1 2 3 4 14
1 2 3 4 8 +(1 Z ijk )Z ijk Z ijk Z ijkC ijk
1
+Z ijk 2
Z ijk 3
Z ijk 4
Z ijk 15
C ijk }x ijk
1
+Z ijk 2
(1 Z ijk 3
)Z ijk 1 Z i4jk C ijk
9
( )
1
Z ijk 2
+ Z ijk 3
+ Z ijk 4
+ Z ijk 1
1
+Z ijk 2
Z ijk (3
1 Z ijk )(1 Z )C 4
ijk
10
ijk

1
+Z ijk 2
(1 Z ijk 3
)Z ijk Z ij4kC ijk
11 x ijk 0, i , j and k
1
+Z ijk 3
(1 Z ijk 2
)Z ijk 4
Z ijk 12
C ijk
q
Z ijk = 0 / 1,q = 1, 2, 3, 4

1
+Z ijk 2
Z ijk 3
Z ijk 4
1 Z ijk (
C i13jk )
1
+(1 Z ijk 2
)Z ijk 3
Z ijk 4
Z ijk 14
C ijk
1 2 We rewrite the objective function (2) as
+(1 Z ijk )(1 Z ijk )
follows:
(1 Z )(1 Z
3
ijk
4
ijk
15
)C ijk }x ijk
m n s
1
Z ijk 2
+ Z ijk 3
+ Z ijk 4
+ Z ijk 3 min :Z =
i =1 j =1 k =1
{C 1
ijk
2
,C , .,C
ijk
16
ijk ijk }x
Z q
ijk
= 0 / 1,q = 1, 2, 3, 4
The cost coefficients of the objective Step 16: When p = 17

function have sixteen choices as
{ 1
C ijk 2
,C ijk ,
16
, ,C ijk }
out of which one is to We rewrite the objective function (2) as
be selected. Since the choices have satis- follows:
fied the norms as 2 4 = 16, so there is no
involvement of auxiliary and additional m n s
min :Z =
constraints. Hence we need four binary
i =1 j =1 k =1
variables and they are Z ijk
1 2 3
, Z ijk
4
.
,Z ijk ,Z ijk
{ 1
C ijk 2
,C ijk 17
, ,C ijk x ijk }
Finally we formulate the above objective
function is as follows:
m n s
In this case, seventeen C ijk
1 2
,C ijk 17
{
, ,C ijk }
are considered as the coefficients of the
min :Z =
i =1 j =1 k =1
objective function in the specified multi-
choice programming problem, out of which
1
{(1 Z ijk 2
)(1 Z ijk 3
) 1 Z ijk ( ) one is to be selected. Since, 24< 17<2 5, so
(1 Z )C + Z
4
ijk
1
ijk
1
ijk we need five binary variables as: Z ijk
1 2
,Z ijk 3
,Z ijk ,
(1 Z )(1 Z )(1 Z )C
2
ijk
3
ijk
4
ijk
2
ijk

4
Z ijk 5
,Z ijk . We express the 25 = 5C0 + 5C1+5C2+
+(1 Z )Z (1 Z ) (1 Z )C
1 2 3 4 3
5
C3+ 5 C4+5 C5, in which the set of consecu-
ijk ijk ijk ijk ijk
tive binomial coefficients terms as {5C2,
+(1 Z )(1 Z )Z (1 Z )C C3}, whose sum is (10 + 10) = 20. It is just
1 2 3 4 4 5
ijk ijk ijk ijk ijk
greater than the number of goals as p = 17.
+(1 Z )(1 Z ) (1 Z ) Z C
1
ijk
2
ijk
3
ijk
4
ijk
5
ijk The auxiliary constraints are also depend on
the ranges of above said of binomial coef-
+Z Z (1 Z ) (1 Z )C
1 2 3 4 6
ijk ijk
ijk ijk ijk ficients. So there are three (2017 = 3) terms
1 2 4 7 3
+Z (1 Z ijk
ijk
)Z (1 Z ijk )C ijk ijk of additional constraints are appeared in the
possible model which contain altogether 120
1
+(1 Z ijk 2
)Z ijk 3
Z ijk 4
1 Z ijk 9
C ijk ( ) similar models. Here only one model is
1 8
+Z (1 Z )(1 Z )Z C ijk
ijk
2
ijk
3
ijk
4
ijk
shown below, as other models are similar:
1 2 3 4 10
+(1 Z ijk )Z ijk (1 Z ijk )Z ijkC ijk m n s
1 2 3 4 11
min :Z =
+(1 Z ijk )(1 Z ijk )Z ijk Z ijkC ijk
i =1 j =1 k =1
1
+Z Z Z (1 Z )C
ijk
2
ijk
3
ijk
4
ijk
12
ijk

1
ijk
3
(1 Z )(1 Z ) 1 Z ijk 2
ijk
4
Z ijk (5
Z ijk 1
C ijk )
1 2 3
1
+Z Z (1 Z )Z C
ijk
2
ijk
3
ijk
4
ijk
13
ij k
+(1 Z ) (1 Z )Z
ijk ijk ijk
1
+Z ijk ( 2
1 Z ijk 3
Z ijk 4
Z ijk 14
C ijk )
(1 Z ) Z
4
ijk
5
ijk
2
C ijk 1
+ (1 Z ijk )

1
+(1 Z ijk 2
)Z ijk 3
Z ijk 4
Z ijkC i15jk
2
(1 Z ijk 3
)Z ijk 4
Z ijk (5
1 Z ijk 3
C ijk )
1
+Z ijk 2
Z ijk 3
Z ijk 4
Z ijk 16
C ijk }x ijk 1
+(1 Z ijk (3
) 1 Z ijk 4
1 Z ijk)( 2
Z ijk ) 5
Z ijk 4
C ijk
+Z ijk 4
Z ijk ( 1
1 Z ijk )(1 Z ) 3
ijk

5 5 1 3 2
q
Z ijk = 0 / 1,q = 1, 2, 3, 4 (1 Z ijk )C ijk + (1 Z ijk )Z Z ijk ijk
out of which one is to be selected. Since, 24<

4
(1 Z ijk 5
) 1 Z ijk( 6
C ijk ) 18<25, so we need five binary variables as:
1
+ Z ijk 2
(1 Z ijk 3
)(1 Z ijk 4
) 1 Z ijk 5
Z ijk 7
C ijk ( ) 1
Z ijk 2
,Z ijk 3
,Z ijk 4
, Z ijk 5
,Z ijk . We express 18 is just
less than as {5C2, 5C3} [whose sum is equal to
1
+Z ijk 2
(1 Z ijk 3
)(1 Z ijk 4
)Z ijk 5
1 Z ijk 8
C ijk ( ) 20] in which binomial coefficients of its
1
+Z ijk 2
(1 Z ijk 3
)Z ijk 4
1 Z ijk ( )(1 Z )C 5
ijk
9
ijk
ranges are established for obtaining the auxil-
iary constraints. Then two (20 18 = 2) addi-
1
+Z ijk 2
Z ijk 3
1 Z ijk( )(1 Z 4
ijk )(1 Z )C 5
ijk
10
ijk
tional constraints are appear in the possible
models. In this case, there are forty five similar
1
+(1 Z ijk 2
)(1 Z ijk 3
)Z ijk 4
Z ijk 5
Z ijk 11
C ijk models are possible. Here only one model is
1 3 2 4 5 12
shown below:
+(1 Z ijk )(1 Z ijk
k
)Z ijk Z ijk Z ijkC ijk
1
+(1 Z ijk 2
)Z ijk 3
Z ijk 4
1 Z ijk 5
Z ijk 13
(
C ijk ) min :Z =
m n s
1
+(1 Z ijk 2
)Z ijk Z ij3jk Z ijk
4 5
1 Z ijk (
C ijk ) 14
i =1 j =1 k =1
1
+Z ijk 2
(1 Z ijk 3
) 1 Z ijk (4
Z ijk 5
Z ijk 15
C ijk )
1
{(1 Z ijk 2
)(1 Z ijk 3
) 1 Z ijk ( )
4 5 1 1 2 3
Z Z C
ijk ijk ijk
+ (1 Z ) (1 Z )Z ijk ijk ijk
1
+Z (1 Z )Z
ijk
2
ijk
3
ijk (1 Z ) Z 4
ijk
5
ijk
C
16
ijk
(1 Z ) Z 4 5 2
C ijk 1
+ (1 Z ijk )
ijk ijk

1
+Z ijk 2
(1 Z ijk 3
)Z ijk (1 Z ) Z 5
ijk
4
ijk
17
C ijk }x ijk 2
(1 Z ijk 3
)Z ijk 4
Z ijk 5
1 Z ijk 3
C ijk ( )
3 ( )( )
1 2 3 4 5
2 Z ijk + Z ijk + Z ijk + Z ijk +Z ijk 1
+(1 Z ijk 3
) 1 Z ijk 4
1 Z ijk 2
Z ijk 5
Z ijk 4
C ijk
Z 1
+Z 2
+Z 5
2
ijk ijk ijk
2
+Z ijk 4
Z ijk 1
1 Z ijk ( )(1 Z ) 3
ijk

1
Z ijk 2
+ Z ijk 4
+ Z ijk 2 5
(1 Z ijk 5
)C ijk 1
+ (1 Z ijk 3
)Z Z ijk 2
ijk
1
Z ijk 2
+ Z ijk 3
+ Z ijk 2 4
(1 Z ijk ) 5
) 1 Z ijk ( 6
C ijk 1
+ Z ijk

x ijk 0, i , j and k (1 Z )(1 Z ) (1 Z ) Z C
2
ijk
3
ijk
4
ijk
5
ijk
7
ijk
q
Z ijk = 0 / 1,q = 1, 2, 3, 4, 5 +Z (1 Z )(1 Z )Z (1 Z )C
1
ijk
2
ijk
3
ijk
4
ijk
5
ijk
8
ijk

+Z (1 Z )Z (1 Z ) (1 Z )C
1
ijk
2
ijk
3
ijk
4
ijk
5
ijk
9
ijk

+Z Z (1 Z ) (1 Z ) (1 Z )C
1
ijk
2
ijk
3
ijk
4
ijk
5
ijk
10
ijk

We rewrite the objective function (2) as 1 2 3 4 5 11
follows: +(1 Z ijk )(1 Z ijk )Z ijk Z ijk Z ijkC ijk
1 3 2 4 5 12

+(1 Z ijk )(1 Z ijk
k
m n s
min :Z =
1
+(1 Z ijk 2
)Z ijk 3
Z ijk 4
1 Z ijk 5
Z ijk 13
C ijk ( )

i =1 j =1 k =1
+(1 Z )Z Z Z 1 2 3
(1 Z )C 4 5 14
{C 1
ijk
2
,C , ,C
ijk
18
ijk }x
ijk
1
+Z ijk 2
(1 Z ijk ) (1 Z ) Z Z C 3
ijk
4
ijk
5
ijk
15
ijk

The cost coefficients of the objective func- 1

+Z ijk 2
(1 Z ijk )Z (1 Z ) Z C
3
ijk
4
ijk
5
ijk
16
ijk

tion have eighteen choices C ijk 1 2
,C ijk 18
, ,C ijk { }
1
+Z ijk 2
(1 Z ijk 3
)Z ijk 5
1 Z ijk 4
Z ijk (17
C ijk ) (1 Z ) Z 4
ijk
5
ijk
2
C ijk 1
+ (1 Z ijk )

1
+Z (1 Z )Z 3 2
(1 Z ) Z 4 5
C }x ijk
18
2
(1 Z ijk 3
)Z ijk 4
Z ijk 5
1 Z ijk( 3
C ijk )
1
2 Z ijk 2
+ Z ijk 3
+ Z ijk 4
+ Z ijk 5
+Z ijk 3
1
+(1 Z ijk (
3
) 1 Z ijk 4
1 Z ijk )(
2
Z ijk 5
)
Z ijk 4
C ijk
1
Z ijk 2
+ Z ijk 4
+ Z ijk 2
2
+Z ijk 4
Z ijk ( 1
1 Z ijk )(1 Z ) 3
ijk

5 5 1 3 2
(1 Z ijk )C ijk + (1 Z ijk )Z Z ijk ijk
Z 1
ijk
+Z 2
ijk
+Z 3
ijk
2
4
(1 Z ijk ) ( 5
) 1 Z ijk 6
C ijk 1
+ Z ijk

(1 Z )(1 Z ) (1 Z ) Z C
2
ijk
3
ijk
4
ijk
5
ijk
7
ijk
Z q
= 0 / 1,q = 1, 2, 3, 4, 5
+Z (1 Z )(1 Z )Z (1 Z )C
ijk
1
ijk
2
ijk
3
ijk
4
ijk
5
ijk
8
ijk

Step 18: When p = 19 +Z (1 Z )Z (1 Z ) (1 Z )C

1
ijk
2
ijk
3
ijk
4
ijk
5
ijk
9
ijk

We rewrite the objective function (2) as +Z Z (1 Z ) (1 Z ) (1 Z )C

1
ijk
2
ijk
3
ijk
4
ijk
5
ijk
10
ijk

follows: 1 2 3 4 5 11
+(1 Z ijk )(1 Z ijk )Z ijk Z ijk Z ijkC ijk
1 3 2 4 5 12

m n s +(1 Z ijk )(1 Z ijk
k
min :Z =
i =1 j =1 k =1 1
+(1 Z ijk 2
)Z ijk 3
Z ijk 4
1 Z ijk 5
(
Z ijk 13
C ijk )
{C 1
ijk
2
,C , ,C
ijk
19
ijk }x
ijk
1
+(1 Z ijk 2
)Z ijk Z ij3jk Z ijk
4 5
1 Z ijk ( 14
C ijk )
1
+Z ijk 2
(1 Z ijk 3
(
) 1 Z ijk 4
Z ijk 5
Z ijk 15
C ijk )
tion have nineteen choices C ijk 1 2
,C ijk 19
, ,C ijk { } 1
+Z ijk 2
(1 Z ijk 3
)Z ijk 4
1 Z ijk (
5
Z ijk 16
C ijk )
out of which one is to be selected. Since, 2 < 4
19<25, so we need five binary variables as:

1
+Z ijk 2
(1 Z ijk 3
)Z ijk 4
Z ijk 5
1 Z ijk (
17
C ijk )
1 2 3 4 5
. We express 18 is just
Z ijk ,Z ijk ,Z ijk , Z ijk ,Z ijk 1
+Z ijk 2
Z ijk 3
(1 Z ijk 4
) 1 Z ijk (5
Z ijk 18
C ijk )
less than as { C2, C3} [whose sum is equal to
5 5
20] in which binomial coefficients of its 1

+Z ijk 2
Z ijk 3
(1 Z ijk 4
)Z ijk 5
1 Z ijk (
19
C ijk }x ijk )
ranges are established for obtaining the auxil-
iary constraints. Then two (20 19 = 1) addi-
1
2 Z ijk 2
+ Z ijk 3
+ Z ijk 4
+ Z ijk 5
+Z ijk 3
tional constraints are appear in the possible
models. In this case, there are forty five similar
1
Z ijk 2
+ Z ijk 3
+ Z ijk 2
models are possible. Here only one model is x ijk 0, i , j and k
shown below:
q
Z ijk = 0 / 1,q = 1, 2, 3, 4, 5
m n s
min :Z =
i =1 j =1 k =1 Step 19: When p = 20
1 3
{(1 Z )(1 Z ) 1 Z ijk
ijk
4
Z ijk2
Z ij5kC ijk
ijk
1
( )
1 3
+(1 Z ) (1 Z )Z ijk 2 We rewrite the objective function (2) as
ijk ijk
follows:
min :Z =
m n s 1
+Z ijk 2
(1 Z ijk ( 3
) 1 Z ijk )
4
Z ijk 5
Z ijk 15
C ijk

{ 1
C ijk 2
,C ijk
i =1 j =1 k =1
20
, ,C ijk x ijk }
1
+Z ijk 2
(1 Z ijk 3
)Z ijk ( 4
1 Z ijk 5
Z ijk )
16
C ijk

1
+ Z (1 Z )Z
ijk
2
ijk
3
ijk(1 Z ) Z C
5
ijk
4
ijk
17
ijk
We state here twenty choices as

1
+Z ijk 3
(1 Z ijk 2
)Z ijk (1 Z ) Z C
4
ijk
5
ijk
18
ijk
{ 1
C ijk 2
,C ijk 20
, ,C ijk }
out of which one is to be 1
+Z ijk 3
(1 Z ijk 2
)Z ijk (1 Z ) Z C
5
ijk
3
ijk
19
ijk
selected. Since, 24< 20<25, so we need five
binary variables as: Z ijk
1 2
,Z ijk 3
,Z ijk 4
, Z ijk 5
,Z ijk . 1
+Z ijk 4
(1 Z ijk 2
)Z ijk (1 Z ) Z C }x
5
ijk
3
ijk
20
ijk ijk

We express the consecutive binomial coef-
ficients as {5C2, 5C3} whose sum is 20 which
1
2 Z ijk 2
+ Z ijk 3
+ Z ijk 4
+ Z ijk 5
+Z ijk 3
is exactly equal to the number of goal as p x ijk 0, i , j and k
= 20. So there does not arise any additional
constraint in such model which is given q
Z ijk = 0 / 1,q = 1, 2, 3, 4, 5
below:
m n s
min :Z = NUMERICAL EXAMPLE

i =1 j =1 k =1
A reputed betel leaves (Pan) supplier company
1
{(1 = Z ijk 3
)(1 Z ) 1 Z ijk 2 4
Z ijk
ijk (
Z ij5kC ijk
1
) transports a variety of Betel leaves from three
1 3 2
+(1 Z ijk ) (1 Z )Z ijk ijk source points namely Ramnagar of Purba
Midnapore, Mecheda of Purba Midnapore
(1 Z ) Z 4
ijk
5
ijk
2
C ijk 1
+ (1 Z ijk )
and Jahalda of Paschim Midnapore in West
2
(1 Z ijk 3
)Z ijk 4
Z ijk 5
1 Z ijk ( 3
C ijk ) Bengal, India by truck or goods train to the four
destination centers at Delhi, Patna, Ranchi and
1
+(1 Z ijk 3
(
) 1 Z ijk 4
1 Z ijk )(
2
Z ijk 5
Z ijk) 4
C ijk Mumbai, India through the eighteen routes. The
transportation cost is related with fluctuation
2
+Z ijk 4
Z ijk (1
1 Z ijk )(1 Z ) 3
ijk
in road condition, distance of destination point
5
(1 Z ijk 5
)C ijk 1
+ (1 Z ijk 3
)Z Z ijk 2
ijk
from source point, consumption of fuel etc. The
main purpose is to minimize the transportation
4
(1 Z ijk ) ( 5
) 1 Z ijk 6
C ijk 1
+ Z ijk

cost and maximize the profit against the market
(1 Z )(1 Z ) (1 Z ) Z C
2 3 4 5 7 price at different markets. The transportation
ijk ijk ijk ijk ijk
cost of carrying one unit (100 Kg) of Betel
+Z (1 Z )(1 Z )Z (1 Z )C
1
ijk
2
ijk
3
ijk
4
ijk
5
ijk
8
ijk
leaves from sources to destinations each treated
as of multi-choice parameters.
+Z (1 Z )Z (1 Z ) (1 Z )C
1
ijk
2
ijk
3
ijk
4
ijk
5
ijk
9
ijk
Without using the multi-choice program-
ming methodology, the problem cannot be
+Z Z (1 Z ) (1 Z ) (1 Z )C
1
ijk
2
ijk
3
ijk
4
ijk
5
ijk
10
ijk
solved easily. Due to increasing the fuel price
rate and road collection tax, the price rates of
1 2 3 4 5 11
+(1 Z ijk )(1 Z ijk )Z ijk Z ijk Z ijkC ijk transportation costs in each route are appended

1
+(1 Z ijk 3
)(1 Z ijk 2
)Z ijk 4
Z ijk 5
Z ijk 12
C ijk below:
k
1
+(1 Z ijk 2
)Z ijk 3
Z ijk 4
1 Z ijk 5
Z ijk( 13
C ijk ) x111 route either 10 or 11 or 12 required
admissible costs per hundred Rupees.
1
+(1 Z ijk 2
)Z ijk 3
Z ijk 4
Z ijk 5
1 Z ijk 14
C ijk (
)
x112 route either 15 or16 required admissible We formulate a MCSSTP where the objec-
costs per hundred Rupees. tive function and the constraints are as follows:
x121 route either 18 or 19 or 20 or 21 required
admissible costs per hundred Rupees. 3 3 3
x122 route either 13 or 14 or 15 required min : Z = C ijk

1 2
,C ijk { p
,...,C ijk }
admissible costs per hundred Rupees. i =1 j =1 k =1
x131 route either 12 or 13 required admis- x ijk , p = 1, 2, 3, 4

sible costs per hundred Rupees.
x132 route either 10 or 14 required admis- Subject to:
x211 route either 9 or 10 or 11 required
admissible costs per hundred Rupees. 3 2
Pr x ijk ai 1
x212 route either 14 or 15 required admis- j =1 k =1
sible costs per hundred Rupees. i , i = 1, 2, 3
x221 route either 22 or 24 required admis-
sible costs per hundred Rupees. 3 2
Pr x ijk bj 1
x222 route either 24 or 25 required admis- i =1 k =1
sible costs per hundred Rupees. j , j = 1, 2, 3
admissible costs per hundred Rupees. 3 3
Pr x ijk ek 1
x232 route either 15 or 20 required admis- i =1 j =1
sible costs per hundred Rupees. k , k = 1, 2
admissible costs per hundred Rupees. x ijk 0 , i , j and k
x312 route either 19 or 20 required admis-
x321 route either 17 or 18 required admis- Due to conflicting situation of market, assume
sible costs per hundred Rupees. that the parameters supplies (sources) ai, demands
x322 route either 12 or 14 or 16 required (destinations)bj andconveyances(capacities)ek ofour
admissible costs per hundred Rupees. specified problem followed by Cauchys distribution
x331 route either 18 or 19 required admis- with scale and shape parameters are given in Tables
sible costs per hundred Rupees. 1-3: Taking two known parameters of Cauchys dis-
x332 route either 14 or 16 required admis- tribution with specified probability level of supplies
sible costs per hundred Rupees. i.e, ai for i = 1, 2, 3 are represented in Table 1.
Table 1. The values of location and scale parameters with SPL for supply
Location Parameter Scale Parameter Specified Probability Level (SPL)
1 = 900 1 = 6.3 1 = 0.09
2 = 1000 2 = 6.9 2 = 0.08
3 = 1100 3 = 7.5 1 = 0.07
Table 2. The values of location and scale parameters with SPL for demand
1 = 600 1 = 4.5 1 = 0.06
2 = 700 2 = 5.0 2 = 0.05
3 = 800 3 = 5.7 3 = 0.04
Table 3. The values of location and scale parameters with SPL for conveyance
1 = 300 1 = 2.0 1 = 0.03
2 = 400 2 = 2.7 2 = 0.02
Taking two known parameters of Cauchys 3 2

distribution with specified probability level of x 2jk
999.8409399 (28)
Demands. i.e, bj for j = 1, 2, 3 are represented j=1 k=1
in Table 2. 3 2
Taking two known parameters of Cauchys x 3jk

1099.822992 (29)
j=1 k=1
distribution with specified probability level of
conveyances i.e, ek for k = 1, 2 are represented 3 2
in Table 3. x i1k
600.1086746 (30)
i=1 k=1
Using the data from Tables 1, 2 and 3, the 3 2
multi-choice deterministic solid transportation
problem shown in Box 1 is constructed.
x i2k
700.1234938 (31)
i=1 k=1
Box 1 is subject to: 3 2
x i3k
800.1439115 (32)
3 2 i=1 k=1
x 1jk
899.8582291 (27)
j=1 k=1
Box 1.
min : Z = {10, 11, 12}x111 + {15, 16}x112 + {18, 19, 20, 21}x121
+ {13, 14, 15}x122 +{12, 13}x131 + {10, 14}x132 + {9, 10, 11}x211 + {14, 15}x 212
+ {22, 24}x221 +{24, 25}x 222 + {20, 21, 22}x231 + {15, 20}x 232 + {16, 17, 18}x 311
+ {19, 20}x 312 +{17, 18}x 321 + {12, 14, 16}x 322 + {18, 19}x 331 + {14, 16}x 332
3 3 1
t232 = 15Z 322 1
+ 20(1 - Z 322 )
x ij1 300.051593 (33)
1 2 1 2
i=1 j=1
t311 = 16Z 311Z 311 + 17Z 311 (1 - Z 311 )
3 3 1 2

+18(1 - Z 311 )Z 311
x ij2 400.0711325 (34)
i=1 j=1 1
t312 = 19Z 312 1
+ 20(1 - Z 312 )
t321 = 17Z 321 1
+ 18(1 - Z 321 )
1 2 1 2
t322 = 12Z 322 Z 322 + 14Z 322 (1 - Z 322 )
Now using a new transformation technique,
1 2
we obtain the multi-choice deterministic solid +16(1 - Z 322 )Z 322
transportation problem shown in Box 2. 1
t331 = 18Z 331 1
+ 19(1 - Z 331 )
Box 2 is subject to (27) to (34):
1
t332 = 14Z 332 1
+ 16(1 - Z 332 )
1 2 1 2
t111 = 10Z 111Z 111 + 11Z 111 (1 - Z 111 )

1
1 Z 111 2
+ Z 111 2
1 2
+12(1 - Z )Z 111 111 1
1 Z 122 2
+ Z 122 2
1
t112 = 15Z 112 1
+ 16(1 - Z 112 )
1
1 Z 211 2
+ Z 211 2
1 2 1 2
t121 = 18Z 221Z 221 + 19Z 221 (1 - Z 221 )
1 2 1 2
1
1 Z 231 2
+ Z 231 2
+20(1 - Z 221 )Z 221 + 21(1 - Z 221 )(1 - Z 221 )
1
t122 = 13Z 212 2
Z 212
1
1 Z 311 2
+ Z 311 2

1
+14(1 - Z 212 1
) + 15(1 - Z 212 1
)Z 212 1
1 Z 322 2
+ Z 322 2
1
t131 = 12Z 311 1
+ 13(1 - Z 311 ) x ijk 0, i , j and k Z ijk
q
= 0 / 1 ; q=1,2
1
t132 = 10Z 123 1
+ 14(1 - Z 123 )
1 1 2
t211 = 9Z 121 + 10z 121 (1 - Z 121 )

1
+11(1 - Z 121 2
)Z 121 RESULT AND DISCUSSION
1
t212 = 14Z 122 1
+ 15(1 - Z 122 ) The above problem is solved by Lingo10 pack-
age and to obtain the optimal solution and the
1
t221 = 22Z 221 1
+ 24(1 - Z 221 ) optimal value of the objective function. They are
1
t222 = 24Z 222 1
+ 25(1 - Z 222 ) as: x132= 800.1439, x211=700.1235, and rest of the
decision variables are zero, and the minimum
1
t231 = 20Z 321 1
+ 21z 321 2
(1 - Z 321 ) cost of the objective function is 15002.67. The

1
+22(1 - Z 321 2
)Z 321 optimal value for multi-choice cost coefficient
Box 2.
min : Z = t111 x111 + t112 x112 + t121 x121 + t122 x122 + t131 x1331 + t132 x132
+t211 x 211 + t212 x 212 + t221 x 221 + t222 x 222 + t231 x 231 + t232 x 232
+t311 x 311 + t312 x 312 + t321 x 321 + t322 x 322 + t331 x 331 + t332 x 332
of xijk to the objective function are 11, 15, 21, consideration of Cauchys distribution. Initially,
15, 12, 10, 10, 14, 22, 24, 22, 15, 18, 19, 17, we have transformed all the stochastic con-
16, 18, 14 per hundred rupees respectively. straints into equivalent deterministic constraints
Introduction of binary variables is an im- by stochastic programming approach. Then we
portant concept in multi-choice programming have introduced a transformation technique
for selection of one choice from the set of which selects only one choice from a set of
multi-choice. To formulate the proposed model multi-choice for each coefficient of objective
in this paper we have required the additional/ function and provides an optimal solution of
auxiliary constraints which contain the binary our model.
variables. The number of binary variables for The transportation phenomena, as a
each choice or goal depends on the relation: manifestation of the complex human social, eco-
nomic, and political interactions, are filled with
ln(p) uncertainties. To handle the first uncertainty we
have assumed that the cost coefficients of the
ln(2)
objective function are of multi-choice rather
than of single choice; and the second uncertainty
where p is the number of choice or goal. When involves supply, demand and capacity which are
p = 2, or 4, or 8 or 16 there is no auxiliary all followed by Cauchys distribution. Finally,
constraint to construct of our proposed model. we have concluded that the formulated model
When p = 3, or 7, or 11, or 15 only one auxiliary is highly applicable for real-life solid trans-
constraint is required which contain the binary portation problem and solving this model, the
variable for construction of our model. When p decision maker has provided more information
= 5, or 6, or 9, or 10, or 13, or 14, or 19, or 20 for taking the right decision. A further study is
then there is needed for two or three auxiliary also needed for MCSSTP under multi-objective
constraints respectively. Again when p = 12, environment.
or 18 then there are four auxiliary constraints
respectively. Specially, when p = 17, there are
lot of models each of which contain five binary ACKNOWLEDGMENT
variables and five auxiliary constraints. To
The authors would like to thank the anonymous
reduce the length of the paper, here we have
reviewers and the editor for their insightful
considered only one model instead of different
comments and suggestions.
models in this case. Any model can be consid-
ered for optimal solution. Depending on the
number of choices or goals, we get different
REFERENCES
models of our proposed problem and solving
all these of choices or goals, we get different Arsham, H. (2012). Foundation of linear program-
models of our proposed problem and solving ming: A managerial perspective from solving
all these models, we have obtained the same system of inequalities to software implementation.
optimal solution. Hence it is necessary to solve International Journal of Strategic Decision Sciences,
3(3), 40-60.
only one model.
Biswal, M. P., & Acharya, S. (2009). Transforma-
tion of a multi-choice linear programming problem.
CONCLUSION AND FUTURE Applied Mathematics and Computation, 210(1),
RESEARCH DIRECTIONS 182188. doi:10.1016/j.amc.2008.12.080
Biswal, M. P., & Acharya, S. (2013). Multi-choice
The main aim of this paper is to present the multi-objective mathematical programming model
solution procedure for multi-choice stochastic for integrated production planning: A case study.
solid unbalanced transportation problem with International Journal of Systems Science, 44(9),
16511665. doi:10.1080/00207721.2012.669862
Bit, A. K., Biswal, M. P., & Alam, S. S. (1993). Mahapatra, D. R., Roy, S. K., & Biswal, M. P.
Fuzzy programming approach to multi-objective (2013). Multi-choice stochastic transportation
solid transportation problem. Fuzzy Sets and Systems, problem involving extreme value distribution. Ap-
57(2), 183194. doi:10.1016/0165-0114(93)90158-E plied Mathematical Modelling, 37(4), 22302240.
doi:10.1016/j.apm.2012.04.024
Chang, C.-T. (2007). Multi-choice goal program-
ming, OMEGA. International Journal of Manage- Maity, G., & Roy, S. K. (2014). Solving multi-choice
ment Sciences, 35, 389396. multi-objective transportation problem: A utility func-
tion approach. Journal of Uncertainty Analysis and
Chang, C.-T. (2008). Revised multi-choice goal Applications, 2(1), 11. doi:10.1186/2195-5468-2-11
programming. Applied Mathematical Modelling,
32(12), 25872595. doi:10.1016/j.apm.2007.09.008 Messaoudi, L., & Reba, A. (2014). A fuzzy goal
programming approach for solving stochastic
Chang, C.-T., Ku, C.-Y., & Ho, H.-P. (2010). Fuzzy portfolio selection problem in a group decision-
multi-choice goal programming for supplier selec- making context. International Journal of Applied
tion. International Journal of Operations Research Decision Sciences, 7(2), 208221. doi:10.1504/
and Information Systems, 1(3), 2851. doi:10.4018/ IJADS.2014.060331
joris.2010070103
Patel, G., & Tripathy, J. (1989). The solid transporta-
Dantzig, G. B. (1963). Linear programming and tion problem and its variants. International Journal
extensions. Princeton, NJ: Princeton University Press. of Management and Systems, 5, 1736.
Goicoechea, A., Hansen, D. R., & Duckstein, L. Rabindran, A., Philips Don, T., & Solberg James, J.
(1982). Multi-objective decision analysis with en- (1987). Operations research: Principles and practice
gineering and business application. New York, NY: (2nd ed.). New York, NY: John Wiley and Sons.
John Wiley and Sons.
Rao, S. S. (1984). Optimization theory and applications
Haley, K. B. (1962). The solid transportation problem. (2nd ed.). New Delhi, India: Wiley Eastern Limited.
Operations Research, 11(4), 448463. doi:10.1287/
opre.10.4.448 Roy, S. K. (2014). Multi-choice stochastic trans-
portation problem involving Weibull distribution.
Hiller, F., & Lieberman, G. (1990). Introduction to International Journal of Operational Research vol
operations research. New York, NY: Mc Graw- Hill. 21(1), 38-58.
Hitchcock, F. L. (1941). The distribution of a product Roy, S. K., & Mahapatra, D. R. (2011). Multi-
from several sources to numerous localities. Journal objective interval valued transportation prabab-
of Mathematics and Physics, 20, 224230. listic problem involving log-normal. International
Jimenez, F., & Verdegay, J. L. (1998). Uncertain Journal of Mathematics and Scientific Computing,
solid transportation problems. Fuzzy Sets and 1(2), 1421.
Systems, 100(1-3), 4557. doi:10.1016/S0165- Roy, S. K., Mahapatra, D. R., & Biswal, M. P. (2012).
0114(97)00164-4 Multi-choice stochastic transportation problem
Liu, S. T. (2006). Fuzzy total transportation cost with exponential distribution. Journal of Uncertain
measures for fuzzy solid transportation problem. Systems, 6(3), 200213.
Applied Mathematics and Computation, 174(2), Shell, E. (1955). Distribution of a product by several
927941. doi:10.1016/j.amc.2005.05.018 properties, directorate of management analysis. In
Mahapatra, D. R., Roy, S. K., & Biswal, M. Proceedings of the Second Symposium in Linear
P. (2010). Stochastic based on multi-objective Programming (Vol. 2, pp. 615-642). DCS/Comptrol-
transportation problems involving normal ran- ler H.Q.U.S.A.F., Washington, DC. Jan.
domness. Advanced Modeling and Optimization, Tao, Z., & Xu, J. (2012). A class of rough mul-
12(2), 205223. tiple objective programming and its application to
Mahapatra, D. R., Roy, S. K., & Biswal, M. P. solid transportation problem. Information Sciences,
(2011). Computation of multi-objective probabilistic 188(1), 215-235.
transportation problems involving normal distribu- Wang, J. (2012). Innovations in information systems
tion with joint constraints. The Journal of Fuzzy for business functionality and operations manage-
Mathematics, 19(4), 865876. ment. IGI Global. doi:10.4018/978-1-4666-0933-4
Sankar Kumar Roy is an Associate Professor of Mathematics and Head of the Department of
Applied Mathematics with Oceanology and Computer Programming of Vidyasagar University,
West Bengal, India. He received his Ph. D. in Operations Research at Department of Mathematics,
Indian Institute of Technology, Kharagpur, India in 2003. His research interests are Operations
Research (Transportation Problem, Game Theory), Lattice in Discrete Mathematics, Rough Set
Theory, Fuzzy Set Theory, Stochastic Programming and Multi-Choice Programming. He has
credit of several research papers published in various journals such as Applied Mathematical
Modelling, Journal of Uncertainty Analysis and Applications, The Journal of Fuzzy Mathematics,
Journal of Uncertain Systems, International Journal of Mathematics and Scientific Computing,
Ricerca Operativa, International Journal of Operational Research, CiiT International Journal of
Fuzzy Systems, Advanced Modelling and Optimization, Bulletin of Calcutta Mathematical So-
ciety, Journal of Physical Sciences, Journal of Information and Computing Sciences, Opsearch,
International Journal of Uncertainty, Fuzziness and Knowledge Based Systems.
Deshabrata Roy Mahapatra is a doctoral student at the Department of Applied Mathematics with
Oceanology and Computer Programming of Vidyasagar University, West Bengal, India. He holds
a M.Sc. and B.Sc. in Applied Mathematics from Kalyani University in India. His research interests
are in Transportation Problem, Multi Choice Programming, and Fuzzy Set Theory. Deshabrata
Roy Mahapatra has published in journals such as The Journal of Fuzzy Mathematics, Journal
of Uncertain Systems, International Journal of Mathematics and Scientific Computing Advanced
Modelling and Optimization, Bulletin of Calcutta mathematical Society, Applied Mathematical
Modelling, Journal of Physical Sciences.

Solving Solid Transportation Problem With Multi Choice Cost and Stochastic Supply and Demand PDF

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Solving Solid Transportation Problem With Multi Choice Cost and Stochastic Supply and Demand PDF

Hochgeladen von

Copyright:

Verfügbare Formate

International Journal of Strategic Decision Sciences, 5(3), 1-26, July-September 2014 1

Solving Solid Transportation

Keywords: Cauchys Distribution, Multi-Choice Programming, Solid Transportation Problem, Stochastic

INTRODUCTION each origin to various destinations, satisfying

of customers. He developed a procedure to solve Roy (2014) solved multi-choice multi-objective

Model 1: 2. Only bj (j = 1,2,, n) follows Cauchys

subject to,Pr(x ijk ai ) 1 i Only ai (i = 1, 2, , m) Follows

m s It is assumed that ai (i = 1, 2, , m) are inde-

mined confidence levels are defined as 0 < i i = 1, 2, , m

1. Only ai (i = 1,2,, m) follows Cauchys xijk

(10) f (bj )d (bj ) 1 j ;

Assume that b j (j = 1, 2, , n) are indepen-

Pr(x ijk ek ) 1 k Model 2:

can be reduced to cumulative density function

EQUIVALENT MODEL WITH min :Z = C ijk

min :Z = Model 2(b):

Step 2: When p = 3 x ijk 0, i , j and k

Step 3: When p = 4 The cost coefficients have five choices as

so we construct the objective function in the

Step 4: When p = 5 min :Z =

We rewrite the objective function (2) as

1 2 3 Then we put restriction to remaining two terms

Step 5: When p = 6 We rewrite the objective function (2) as

(8 7) by introducing auxiliary and ad- q

Model 6(a): We present the objective function (2) as

+ (1 Z ) (1 Z ) Z C 1 2 3 3 ijk ijk ijk ijk

Step 8: When p=9 1

The cost coefficients of the objective

iliary and additional constraints in the + (1 Z Z (1 Z )C

we formulate two different models. Again

tion have ten choices as C ijk

of which one is to be selected. Since 23< 10 + (1 Z Z (1 Z )C

In each model, the auxiliary constraints depend

x ijk 0, i , j and k selected. Since 23< 12<24, so we need four

level since p = 12. To reduce the length of m n s

ijk ijk ijk ijk ijk

two sets of consecutive terms of binomial coef- Model 14(b):

auxiliary constraints are performed with the min :Z =

with their restriction as sum of four binary

The cost coefficients of the objective Step 16: When p = 17

out of which one is to be selected. Since, 24<

The cost coefficients of the objective func- 1

Step 18: When p = 19 +Z (1 Z )Z (1 Z ) (1 Z )C

We rewrite the objective function (2) as +Z Z (1 Z ) (1 Z ) (1 Z )C

19<25, so we need five binary variables as:

20] in which binomial coefficients of its 1

We state here twenty choices as

min :Z = NUMERICAL EXAMPLE

x122 route either 13 or 14 or 15 required min : Z = C ijk

x131 route either 12 or 13 required admis- x ijk , p = 1, 2, 3, 4

Location Parameter Scale Parameter Specified Probability Level (SPL)

1 = 900 1 = 6.3 1 = 0.09

2 = 1000 2 = 6.9 2 = 0.08

3 = 1100 3 = 7.5 1 = 0.07

Location Parameter Scale Parameter Specified Probability Level (SPL)

1 = 600 1 = 4.5 1 = 0.06

2 = 700 2 = 5.0 2 = 0.05

3 = 800 3 = 5.7 3 = 0.04

Location Parameter Scale Parameter Specified Probability Level (SPL)

1 = 300 1 = 2.0 1 = 0.03

2 = 400 2 = 2.7 2 = 0.02

Taking two known parameters of Cauchys 3 2

Taking two known parameters of Cauchys x 3jk

subject to,Pr(x ijk ai ) 1 i Only ai (i = 1, 2, , m) Follows