JOURNAL OF TRANSPORTATION

SYSTEMS ENGINEERING AND INFORMATION TECHNOLOGY

Volume 10, Issue 6, December 2010
Online English edition of the Chinese language journal

Cite this article as: J Transpn Sys Eng & IT, 2010, 10(6), 5363.


Received date: Jul 24, 2009; Revised date: Nov 16, 2009; Accepted date: Nov 30, 2009
*Corresponding author. E-mail: [email protected]
Copyright 2010, China Association for Science and Technology. Electronic version published by Elsevier Limited. All rights reserved.
DOI: 10.1016/S1570-6672(09)60071-4
RESEARCH PAPER

Cost-time Minimization in a Transportation Problem
with Fuzzy Parameters: A Case Study
CHAKRABORTY Ananya*, CHAKRABORTY M.

Department of Applied Mathematics, Indian School of Mines, Dhanbad-826004, Jharkhand, India


Abstract: In real world applications the supply, the demand and the transportation cost per unit of the quantities in a transportation
problem are hardly specified precisely because of the changing economic and environmental conditions. It is also important that the
time required for transportation should be minimum. In this paper a method has been proposed for the minimization of transportation
cost as well as time of transportation when the demand, supply and transportation cost per unit of the quantities are fuzzy. The
problem is modeled as multi objective linear programming problem with imprecise parameters. Fuzzy parametric programming has
been used to handle impreciseness and the resulting multi objective problem has been solved by prioritized goal programming
approach. A case study has been made using the proposed approach.

Key Words: transportation economy; impreciseness; fuzzy parametric programming; multiple criteria optimization; goal
programming



1 Introduction
The problem of minimizing the total cost of transportation
has been studied since long and is well known. In a time
minimizing transportation problem, the time of transporting
goods is minimized to satisfy certain conditions in respect of
availabilities at sources and requirements at destinations. The
basic difference between cost minimizing and time
minimizing transportation problem is that the cost of
transportation changes with variations in the quantity but the
time involved remains unchanged and irrespective of the
quantities. The time minimizing transportation problem has
been studied by Hammer
[1,2]
, Garfinkel and Rao
[3]
, Szware
[4]
,
Bhatia, Swarup and Puri
[5]
, Ramakrishnan
[6]
, Sharma and
Swarup
[7]
, Seshan and Tikekar
[8]
and by several other authors.
Time-cost trade off means the problem of minimizing the
transportation cost in addition to minimizing the time of the
transportation. Time-cost trade off analysis has been discussed
by Satya Prakash
[9]
, Bhatia, Swarup and Puri
[10]
and several
other authors. Satya Prakash
[9]
has used goal programming
approach to solve the problem. Liu
[11]
discussed a method for
solving the cost minimization transportation problem with
varying demand and supply.
Most of the models developed for solving the transportation
problem are with the assumption that the supply, demand and
the cost per unit values are exactly known. But in real world
applications, the supply, the demand and the cost per unit of
the quantities are generally not specified precisely i.e. the
parameters are fuzzy in nature. Impreciseness in the
parameters means the information for these parameters are not
complete. But even with incomplete information, the model
user is normally able to give a realistic interval for the
parameters. Carlsson and Korhonen
[12]
and Chanas
[13]

discussed parametric approach to deal with the fuzzy
parameters.
In this paper cost and time minimization transportation
problem has been dealt when the supply, the demand and the
transportation cost per unit of the quantities are fuzzy.
Impreciseness in the parameters has been dealt with using the
concept developed in Ref. [12] and for the time minimization,
the set of times (required for different sources to different
destinations) is partitioned suitably. Linear and exponential
membership functions have been considered for the imprecise
parameters. The problem has been posed as multi objective
linear programming problem and solved using preemptive
goal programming approach by assigning different priorities
CHAKRABORTY Ananya et al. / J Transpn Sys Eng & IT, 2010, 10(6), 5363

to the different objectives. The model intends to obtain a
compromise solution with minimum transportation cost as
well as minimum transportation time. A case study has been
made for the transportation of coal from different collieries of
a regional coal company situated at Jharia coalfield (India) to
different washeries.
2 Formulation of the problem
Suppose that there are m origins and n destinations. It is
assumed that the quantities available at the origins, demand at
the destinations and the cost per unit of the quantity are not
given precisely. However, a realistic interval is available for
each of these imprecise parameters. Further it is assumed that
it is possible to transport the product from any origin to any
destination and the time of the transportation does not depend
on the amount of product transported. Let t
ij
be the units of
time of transportation of the product from origin i to the
destination j, c
ij
be the cost of transportation of one unit of the
product from origin i to the destination j, a
i
be the units of the
product available at origin i, b
j
be the demand at the
destination j and x
ij
be the number of units of the product
transported from origin i to the destination j. The primary
objective is to minimize the total cost of transportation and the
secondary objective is to minimize the duration of the
transportation subject to the constraints.
The fuzzy mathematical formulation of this bi-objective
problem may be presented as follows:
Find x
ij
0 (i =1, 2, , m; j =1, 2, , n) to
Minimize Z
1
=
1 1
ij
m n
ij
i j
c x
= =

%
(1)
Minimize Z
2
=max{t
ij
: x
ij
>0 (i=1, 2, , m; j=1, 2, , n)} (2)
Subject to the constraints:
i
n
j
ij
a x
~

1

=
( i =1, , m) (3)
j
m
i
ij
b x
~

1

=
( j=1, , n) (4)
x
ij
0 (i=1, 2, , m; j=1, 2, , n), where ij
c
~
,
i
a
~
,
j
b
~
are the
imprecise parameters for the transportation cost per unit of
quantity, availability at source and demand at the destinations.
The model applies when the total availability exceeds the total
demand.
These imprecise parameters may be represented by the
intervals [c
ij
0
, c
ij
1
), [a
i
0
, a
i
1
) and [b
j
0
, b
j
1
) respectively. In each
of the intervals the first component with subscript 0 represents
the risk free and the second component with subscript 1
represents the unrealistic/impossible parameters
[12]
. It is
clear that the solution with lower bounds (first component)
should most certainly be implementable and the solution with
upper bound (second component) is practically not
implementable. Thus, when we move from risk free towards
impossible parameter values, we move from solutions with
a high grade to solutions with a low grade on implementing.
i.e., from secure to optimistic solutions. It is required to
find an optimal compromise solution as a function of the
grades of imprecision in the parameters.
3 Solution procedure
There are many possible forms for membership functions:
linear, exponential, hyperbolic, hyperbolic inverse, piece-wise
linear, etc. The linear form of membership function is most
common and suitable to describe impreciseness in many real
world problems. Exponential form of membership functions is
another suitable form, which is not as restrictive as the linear
form, but flexible enough to describe the grades of precision
in the parameter values. Unlike linear membership function,
for nonlinear membership functions the marginal rate of
increase (or decrease) of membership values as a function of
model parameters is not constant. Slowinski
[14]
have used
exponential membership function for the fuzzy multi criteria
linear programming problem. Gupta et al.
[15]
studied a pair of
fuzzy primal-dual linear programming problems and
calculated duality results using an aspiration level approach
using exponential membership function. For the above fuzzy
bi-objective problem, suitable membership functions may be
chosen to match the reality in the field. Based on the analysis
of data, we have considered the linear membership function
for the imprecise availability and the exponential membership
function for the imprecise cost and the demand parameters.
3.1 Multi objective model
Let the membership functions for cost, availability and
demand imprecise parameters are denoted by
i ij
a c
,
and
j
b

, respectively.
We will use the subscript p to refer to a parameter of the
model (1)-(4). Let us scale the membership functions such that

p
=1 if pp
0
and
p
=0 if p p
1
.
The linear membership function (Fig. 1) considered for the
availability, may be defined as
[16]
:
0
1
0 1
0 1
1
1 if
-
if
-
0 if
p
p p
p p
p p p
p p
p p

= < <

(5)


Fig. 1 Linear membership functions
CHAKRABORTY Ananya et al. / J Transpn Sys Eng & IT, 2010, 10(6), 5363

The exponential membership function considered for the
cost and demand parameters may be defined as
[17]
:
)}] /( ) ( exp{ 1 [
1 0 1
p p p p b a
p p p
= (6)
where b
p
>0 or b
p
<0 and a
p
=1/(1exp{b
p
}) and p
e
[ p
0
, p
1
].
The parameter b
p
should be specified by model user (Fig. 2).
The precision of the optimal solution may be evaluated as
follows
[12]
:
The membership function
s
, which outlines the solution,
may be defined as:


Fig. 2 Exponential membership function

min { , , }
ij i j
s c a b
= =
, i=1, 2, , m; j= 1, 2, , n (7)
This means that the precision inherits in the optimal
solution equals the precision of the most risky of the
parameters (values near to upper bound of the intervals). Thus,
in model (1)-(4), the best value for the objective function at
the fixed level of
s
is reached
[12]
when
ij a b
i j
s c
= = =
, for all i=1, 2, , m; j=1, 2, , n (8)
because the model tends to use the risky values of the
parameters.
This can be translated into: the best value of the objective
function, at a fixed level of precision, can be found by using
parameter values of the same level of precision
[12]
. For some
component of the right hand side vector we could probably
use a lower level of precision, because all the constraints are
not active in an optimal solution. This has, however no effect
on the optimal value of the objective function.
Now, to solve the fuzzy optimization problem (1)-(4), we
may proceed as follows:
From Eqs. (5) and (6), we get the parameter p as:
0 1 1
( )
p
p p p p = + (9)
) / 1 log( } / ) {(
1 0 1
p p p
a b p p p p = (10)
On using Eqs. (9) and (10), the fuzzy model (1)-(4) takes
the form
Find x
ij
0 (i = 1, 2, , m; j = 1, 2, , n) to
Minimize Z
1
=
1 0 1
1 1
( ( ))
ij ij ij
m n
ij
i j
c c c x
= =
+

(11)
Minimize Z
2
=max{t
ij
: x
ij
>0 (i=1, 2, , m; j=1, 2, , n)} (12)
Subject to the constraints:
)) ( (
1 0 1
1
i i i
n
j
ij
a a a x +

=

(i=1, 2, ,m) (13)
1 0 1
1
( ( ))
m
ij j j j
i
x b b b
=
+
(j=1, 2, ,n) (14)
x
ij
0 (i=1, 2, , m; j=1, 2, , n), where =log(1/a)/b
(Because
c
=
a
=
b
, we can omit the indices and use in place
of
p
).
In the linear case the parameter : 01 and in the
exponential case : 0log(11/a)/b.
Now, as the membership functions are different in forms, it
is not possible to solve the above optimization problem with
the linear parametric programming approach. To overcome
this difficulty, we may proceed as follows
[12]
:
With different values of membership function =0, 0.1,
0.2, , 1.0 corresponding linear programming problems may
be solved. Then with this set of solutions it is possible to
represent the optimal solutions as functions of the membership
functions and may be presented by graph. The behaviour of
the objective function with varying grades of precision may be
noted in the graph and the appropriate conclusion may be
drawn. The concept may be utilized to get the solution of the
above multi objective problem.
Model (11)-(14) cannot be solved with the classical
approach. An alternate approach is discussed below.
To handle the objective function (12) representing the
minimization of time, we partition the set {t
ij
: i=1, 2, , m; j
=1, 2, , n} into subsets L
k
(k = 1, , q) in the following
way
[9]
.
Each of the subsets L
k
consists of the t
ij
having the same
numerical value. L
1
consists of the t
ij
having the greatest value,
L
2
consists of the t
ij
having the next lower greatest value and
so on. Finally, L
q
consists of the t
ij
having the lowest value.
After obtaining the above q partitioned sets, the above
optimization problem (11)-(14) is transformed to multi
objective linear programming problem with q +1 objectives as
below:
( ) ( )
2
3
1 0 1
1
1 1
2
3
Min
Min
Min

Min
q
m n
ij ij ij ij
i j
ij
L
ij
L
q ij
L
Z c c c x
Z x
Z x
Z x

= =

= +

M
(15)
subject to constraints (13) and (14).
Here
k
ij
L
x

is the sum of amounts x

ij
transported from the
origins to the destinations associated with the t
ij
belonging to
L
k
(k =1, 2, , q).
In the above model it is desired that the transportation cost
CHAKRABORTY Ananya et al. / J Transpn Sys Eng & IT, 2010, 10(6), 5363

is to be minimized to the fullest extent possible, next the sum

1 L
ij
x
of amounts x
ij
transported from the origins to the
destination associated with the t
ij
belonging to L
1
is minimized,
next the sum

2 L
ij
x
is minimized, and so on.
Hence, different priorities may be assigned to the objectives
as per their importance.
Now, the objective function may be constructed as below:
1 2
1 0 1
0
1 1
1 2
Min { ( ( ) ) ,
, ,..., }
q
m n
ij ij ij ij
i j
ij ij q ij
L L L
Z P c c c x
P x P x P x

= =
= +

(16)
subject to the constraints (13) and (14), where P
0
, P
1
, P
2
, , P
q

are the priority levels assigned to different objectives as per
their importance satisfying the relation P
k
<<P
k-1
.
It is to be noted that the given problem is equivalent to the
above alternative problem because minimizing Z given by (16)
determines x
ij0 (i =1, 2, , m; j=1, 2, , n) for which the
total cost and the duration of transportation would be
minimum with priorities being accorded to them in the order
specified above. When we minimize Z given by (16), the total
cost of transportation is minimized to the fullest extent
possible, next the sum of the amounts x
ij
transported from the
origins to the destinations associated with the maximum
values of t
ij
(belonging to L
1
) is minimized, next the sum of
amounts x
ij
transported from the origins to the destinations
associated with the next maximum values of t
ij
(belonging to
L
2
) is minimized and so on.
3.2 Goal programming approach
[18]

The model (16) with (13) and (14) may be converted to a
prioritized goal programming (GP) model. Each of the
objectives may be converted to goals by assigned aspiration
levels. By introducing deviation variables to the constraints
and the goals, a suitable achievement function for the GP
model may be constructed. The resulting prioritized GP model
may be solved using a revised simplex method. The process is
illustrated in section-4 for a case study.
4 Case study
In this section we have considered a transportation problem
to transport raw coal from different collieries of a regional
coal company situated in Jharia (Dhanbad, India) coalfield
area to different washeries around Jharia. In the washeries raw
coal (with comparatively high percentage of ash) are washed
to produce wash coal with low percentage of ash. The washed
coal is frequently used in different industries such as steel
industries, power sectors and other industries. The washed
coal (with different ash percentages) is sent from different
washeries to different industries. The management desires to
transport the raw coal from the different collieries to different
washeries with minimum transportation cost and to supply as
early as possible (i.e. minimum transportation time), so that
the production at the washeries does not hamper.
4.1 Assumptions
We have considered five collieries (of a regional coal
company) and eight washeries (destination points). These
collieries and washeries are situated at reasonable distances. It
is assumed that the washeries are connected by road and road
conditions for all the paths are not well. These collieries have
limited capacities for production of raw coals and the different
washeries have certain demands of raw coal. Due to the
varying reasons and environmental situations the availability
at different collieries, demand at different washeries and
transportation cost per unit are imprecise in nature. However it
is possible to represent these imprecise parameters by realistic
intervals by interacting with the managers. The field data
related to these intervals along with the time of transportation
from different collieries to different washeries is in Table 1.
Table 1 Data representation

Washery
Colliery
W
1
W
2
W
3
W
4
W
5
W
6
W
7
W
8
Availability
a
i
in ton
C
1
Time
[35, 41)
200
[12,14.5)
65
[18,20.6)
100
[22,24.7)
150
[18,20.6)
100
[15, 18)
100
[18,22.8)
100
[22,24.7)
120
[110,146)
C
2

Time
[35, 41)
200
[12, 16)
65
[12,14.5)
65
[21,23.7)
120
[20,22.8)
120
[15, 18)
100
[18,22.8)
100
[20,22.8)
120
[80,120)
C
3

Time
[12, 16)
65
[74,76.2)
300
[35, 41)
200
[22,24.7)
150
[14, 18)
100
[26,28.6)
150
[18,20.6)
100
[8, 10.2)
65
[110,150)
C
4

Time
[12,14.8)
65
[21,24.7)
120
[35, 41)
200
[18,20.6)
100
[12,14.8)
65
[25,28.8)
150
[12, 16)
65
[74,76.2)
300
[130,174)
C
5

Time
[35, 41)
200
[18,22.8)
100
[26,28.6)
150
[20,22.8)
120
[12,15.8)
65
[24,27.6)
150
[15, 18)
100
[64, 66)
300
[100,140)
Demands
b
j
in ton
[80, 112) [70, 90) [60, 84) [60, 92) [70, 106) [60,80.8) [50, 62) [80, 104)
CHAKRABORTY Ananya et al. / J Transpn Sys Eng & IT, 2010, 10(6), 5363

4.2 Data representation
In Table-1, data is presented where time is given in minutes
and transportation cost is given in rupees (Rs). Availability
and demands are in terms of tons. The first component of the
intervals of the availability, demand and cost parameters are
risk free and the second components are risky/unrealistic
values.
4.3 Objectives
The management desires to achieve the following goals:
Goal 1: not to exceed the capacity of availability at the
different collieries but ship out all the materials.
Goal 2: to meet the demands of each washery.
Goal 3: to minimize the total transportation cost.
Goal 4: to minimize the total transportation time.
4.4 Mathematical formulation
For this case the set {t
ij
: i=1, 2, , 5; j=1, 2, , 8} has been
partition into 6 subsets L
k
, k=1, 2, ., 6. The subsets forming
the partition of the set {t
ij
: i=1, 2, , 5; j=1, 2, , 8}are as
follows:
L
1
={t
32
, t
48
, t
58
}
L
2
={t
11
, t
21
, t
33
, t
43
, t
51
}
L
3
={t
14
, t
34
, t
36
, t
46
, t
53
, t
56
}
L
4
={t
18
, t
24
, t
25
, t
28
, t
42
, t
54
}
L
5
={t
13
, t
15
, t
16
, t
17
, t
26
, t
27
, t
35
, t
37
, t
44
, t
52
, t
57
}
L
6
={t
12
, t
22
, t
23
, t
31
, t
38
, t4
1
, t
45
, t
47
, t
55
} (17)
The t
ij
s belonging to L
1
, L
2
, L
3
, L
4
, L
5
and L
6
have the
numerical values 300, 200, 150, 120, 100, 65 respectively.
The four goals considered in section 4.3 may be assigned
different priorities as per their importance. Goal 4 which is
concerned with minimization of time may be divided into
subgoals further and different priorities may be assigned to
these subgoals to achieve the time minimization objective.
Here we have considered goal 4 as minimization of the sum of
maximum time of transportation i.e. concerned to set L
1
and
goal 5 as minimization of weighted sum of next highest time
of transportation i.e. concerned to set L
2
, L
3
, L
4
, L
5
and L
6
.
Now the goals with aspiration levels may be represented as:
Goal 1 (supply goal):
8
1
ij
j
x
=

i
a
~
, i=1, 2, , 5 (18)
Goal 2 (demand goal):

=
5
1 i
ij
x
j
b
~
, j=1, 2, , 8 (19)
Goal 3 (transportation cost goal):

= =
5
1
8
1
~
i
ij
j
ij
x c
0 (20)
Goal 4 (time goal):
x
32
+x
48
+x
58
0 (21)
Goal 5 (Time goal):
11 21 33 43 51
14 34 36 46 53 56
18 24 25 28 42 54
13 15 16 17 26 27 35
37 44 52 57
12 22 23
0
0
0

0

x x x x x
x x x x x x
x x x x x x
x x x x x x x
x x x x
x x x
+ + + + =
+ + + + + =
+ + + + + =
+ + + + + +
+ + + + =
+ +
31 38 41
45 47 55

0
x x x
x x x

+ + +

+ + + =

(22)
Aspiration level 0 has been assigned for transportation cost
goal and time goals. The intervals for
i
a% ,
j
b
%
, and
ij
c%
are
obtained from Table 1.
Based on the interaction with the manager and trend of
variation in the availability, demand and cost, the linear
membership function has been considered for the
impreciseness in the availability and exponential membership
function with b
p
=0.8 and 3.0 having been considered for the
impreciseness in the demand and cost, respectively.
Using (9) and (10), the membership functions for
availability, demand and cost parameters may be written as
1 0 1
1 0 1
1 0 1
( ) / ( )
1 / (1 exp( 0.8))
(1 exp( 0.8 ( ) / ( )))
1 / (1 exp(3.0))
(1 exp(3.0 ( ) / ( )))
i
j
ij
a i i i i
b
j j j j
c
ij ij ij ij
a a a a
b b b b
c c c c

(23)
Consider these membership functions, and the elements of
the model ((a
0
a)/(a
0
a
1
), (b
0
b) / (b
0
b
1
), (c
0
c)/(c
0
c
1
)) may
be represented graphically as shown in Figs. 35.



Fig. 3 The membership function of the parameters of the
availability


Fig. 4 The membership function of the parameters of the demand
for b
p
=0.8
1.0
1.0
1.0
1.0
M
e
m
b
e
r
s
h
i
p

f
u
n
c
t
i
o
n

M
e
m
b
e
r
s
h
i
p

f
u
n
c
t
i
o
n

CHAKRABORTY Ananya et al. / J Transpn Sys Eng & IT, 2010, 10(6), 5363


Fig. 5 The membership function of the parameters of the
transportation cost b
p
=3.0


Now, the parameters can be represented as functions of the
membership functions as follows:
1 0 1
0 1
0 1
( )
log(1 (1 exp( 0.8))) / (0.8( ))
log(1 (1 exp(3.0))) / (3.0( )))
i
j
ij
i i a i i
j j b j j
ij ij c j j
a a a a
b b b b
c c c c

= +

= +

(24)
On introducing the positive and negative deviational
variables, inequations (18)-(21) can be transformed to
equations as:
8
1 1 1
1
8
2 2 2
1
8
3 3 3
1
8
4 4 4
1
8
5 5 5
1
+ - = 146-36
120 40
150 40
174 44
140 40
j
j
j
j
j
j
j
j
j
j
x n p
x n p
x n p
x n p
x n p

=
=
=
=
=

+ =

+ =

+ =

+ =

(25)
5
1 6 6
1
5
2 7 7
1
5
3 8 8
1
5
4 9 9
1
5
5 10 10
1
+ - = 112 - 40 log [1- 0.55 ]
+ - = 90 - 25 log [1- 0.55 ]
+ - = 84 - 30 log [1- 0.55 ]
+ - = 92 - 40 log [1- 0.55 ]
+ - = 106 - 45 log [
i
i
i
i
i
i
i
i
i
i
x n p
x n p
x n p
x n p
x n p

=
=
=
=
=

5
6 11 11
1
5
7 12 12
1
5
8 13 13
1
1- 0.55 ]
+ - = 80.8 - 26 log [1- 0.55 ]
+ - = 62 - 15 log [1- 0.55 ]
+ - = 104 - 30 log [1- 0.55 ]
i
i
i
i
i
i
x n p
x n p
x n p

=
=
=

(26)
ij
j
ij
i
x c

= =
8
1
5
1
+n
14
p
14
=0 (27)
x
32
+x
48
+x
58
+n
15
p
15
=0 (28)
11 21 33 43 51 16 16
14 34 36 46 53 56 17 17
18 24 25 28 42 54 18 18
13 15 16 17 26 27
35 3
- 0
- 0
- 0


x x x x x n p
x x x x x x n p
x x x x x x n p
x x x x x x
x x
+ + + + + =
+ + + + + + =
+ + + + + + =
+ + + + +
+ +
7 44 52 57 19 19
12 22 23 31 38 41 45
47 55 20 20
- 0

- 0
x x x n p
x x x x x x x
x x n p

+ + + + =

+ + + + + +

+ + + =

(29)
where x
ij
0 , n
i
, p
i
0 and 0
i i
n p = , and
c
ij
=c
ij
1
+[(c
ij
0
c
ij
1
)log{1+19.09}]/3.
In the above equations, the deviation variables n
i
and p
i
are
the negative (under achievement) and the positive (over
achievement) deviational variables from the respective goals.
In the above set of equations, the over achievement p
i
(i =1,
2, , 5) for goal-1, under achievement n
i
(i = 6, 7, , 13) for
goal-2, over achievement p
14
for transportation cost goal and
p
i
(i =15, 16, , 20) for goals 4 and 5 are to be minimized.
4.5 Achievement function
The priority levels P
1
, P
2
, P
3
, P
4
and P
5
have been assigned
to the goals 1, 2, 3, 4 and 5, respectively. The priority P
4
has
been assigned to the goal to minimize the maximum duration
of transportation time and the priority P
5
has been assigned to
minimize the rest total transportation time, which is a
weighted sum of deviational variables. These deviational
variables have been kept in the same priority level but
assigned with different cardinal weights to reflect their
relative importance in the priority.
Now with the above preemptive priority structure, the linear
goal programming model may be formulated as:
Minimize
5 13 _
1 2 3 14 4 15
1 6
5 16 17 18 19 20
( , ) [ ( ), ( ), ( ), ( ),
(0.5 0.3 0.15 0.03 0.02 )]
i
i
i i
Z V n p P P P n P p P p
P p p p p p
= =
= =
+ + + +

(30)
Subject to (25)-(29)
5 Numerical solution
Now the above model has become a parametric type goal
programming problem in which cost coefficients, availability
and demand are parameterized. As the membership functions
are of different forms (linear and exponential), it will not be
possible to solve the problem using classical linear parametric
approach. To solve the model, we may solve 11 different
linear goal programming problems with various membership
values
j i ij
b a c
= = =
=0.0, 0.1, , 1.0. On solving this
model with the above different values of the membership
functions, the corresponding values of the objective functions
(transportation cost and transportation time) are obtained. The
value of the decision variables x
ij
, deviation variables and the
achievement of the objective functions are presented in Table
2. The graph of the transportation cost objective function is
also presented in Fig. 6.


M
e
m
b
e
r
s
h
i
p

f
u
n
c
t
i
o
n

1.0
1.0
CHAKRABORTY Ananya et al. / J Transpn Sys Eng & IT, 2010, 10(6), 5363


Table 2 Solutions at different values of
p

Precision level,
p
Decision variables x
ij
Deviation variables n
i
, p
i

Achievement vector
_
Z
0.0 x
12
= 90, x
16
= 56, x
23
= 84,
x
24
= 11.2, x
26
= 24.8, x
31
= 46,
x
38
= 104, x
41
= 66, x
44
= 80,
x
47
= 28, x
55
= 106, x
57
= 34.
p
14
=11 399.24, p
16
= 90,
p
18
= 11.2, p
19
= 194.8,
p
20
= 524.0, n
9
= 0.8.
(0, 0.8, 11 399.24, 0.0, 54.34)
0.1 x
12
= 88.6, x
16
= 53.8, x
23
= 82.32,
x
24
=8.14, x
26
= 25.54, x
31
= 43.68,
x
38
= 102.32, x
41
= 66.08, x
44
= 42.36,
x
47
= 61.16, x
54
= 32.52, x
55
= 103.48.
p
14
=10 283.68, p
16
= 88.6,
p
18
= 40.66, p
19
= 121.7,
p
20
= 547.64, n
9
= 6.74.
(0, 6.74, 10 283.68, 0.0, 65)
0.2 x
12
= 87, x
16
= 51.8, x
23
= 80.4,
x
24
=5.72, x
26
= 25.88, x
31
= 41.6,
x
38
=100.4, x
41
= 65.6, x
44
= 39.6,
x
47
=60.2, x
54
= 31.4, x
55
= 100.6.
p
14
= 9718.68, p
16
= 87,
p
18
= 37.12, p
19
= 117.28,
p
20
= 535.8, n
9
= 10.48.
(0, 10.48, 9 620.863, 0.0, 63.3024)
0.3 x
12
= 85.4, x
16
= 49.8, x
23
= 78.48,
x
24
=3.3, x
26
= 26.22, x
31
= 39.5,
x
38
=98.5, x
41
= 65.1, x
44
= 36.46,
x
47
=59.2, x
54
= 30.28, x
55
= 97.72.
p
14
= 9 084.159, p
16
= 85.4,
p
18
= 33.58, p
19
= 112.48,
p
20
= 523.94, n
9
= 14.56
(0.0, 14.56, 9 084.159, 0.0, 61.59)
0.4 x
12
=83.8, x
16
= 47.8, x
23
= 76.56,
x
24
=.89, x
26
=26.55, x
31
= 37.44,
x
38
=96.56, x
41
=64.64, x
44
= 33.48,
x
47
=58.28, x
54
=29.16, x
55
= 94.84.
p
14
= 8 616.837, p
16
= 83.8,
p
18
= 30.05, p
19
= 107.83,
p
20
= 512.12, n
9
= 18.55.
(0, 18.55, 8 616.837, 0.00, 59.88)
0.5 x
12
=82, x
16
= 46, x
23
= 74.4,
x
26
=25.6, x
31
= 35.6, x
38
= 94.4,
x
41
=63.6, x
44
= 31.2, x
47
=57.2,
x
54
= 28.4, x
55
= 91.6.
p
14
= 8219.107, p
16
= 82,
p
18
= 28.4, p
19
= 102.8,
p
20
= 498.8, n
9
= 19.6.
(0.0, 20.48, 8 219.107, 0.0, 58.32)
0.6 x
12
= 80, x
16
= 44.4, x
23
= 72,
x
26
= 24, x
31
= 34, x
38
= 92,
x
41
= 62, x
44
= 29.6, x
47
= 56,
x
54
= 28, x
55
= 88.
p
14
=7 845.647, p
16
= 80.00,
p
18
= 28, p
19
=98, p
20
= 484
n
9
= 18.4, n
11
= 6.0
(0, 24.4, 7 845.647, 0.00, 56.82)
0.7 x
12
= 55.6, x
16
= 65.2, x
22
= 22.4,
x
23
=69.6, x
31
= 32.4, x
38
= 89.6,
x
41
= 60.4, x
44
= 28, x
47
= 54.8,
x
54
= 27.6, x
55
= 84.4.
p
14
=7 470.204, p
16
= 55.6,
p
18
= 27.6, p
19
=93.2, p
20
= 469.2,
n
9
= 17.2, n
11
= 3.12
(0.0, 20.32, 7 470.204, 0, 44.12)
0.8 x
12
= 54.3, x
16
=62.9, x
22
=21.3,
x
23
=66.7, x
31
=31.28, x
38
= 86.72,
x
41
=57.68, x
44
=27.76, x
47
= 53.36,
x
54
= 27.92, x
55
= 80.08.
p
14
= 7172.645, p
16
= 54.3,
p
18
= 27.92, p
19
= 90.66, p
20
= 451.42,
n
9
= 13.28, n
11
=2.9
(0, 16.18, 7 172.645, 0.00, 43.09)
0.9 x
12
=52.16, x
16
=61.44, x
22
=20.64,
x
23
=63.36, x
31
=30.64, x
38
= 83.36,
x
41
=53.84, x
44
=28.88, x
47
= 51.68,
x
54
= 29, x
55
= 75.
p
14
= 6898.61, p
16
= 52.16,
p
18
= 29, p
19
= 90.32, p
20
= 430.68,
n
9
= 6.62, n
11
=1.46
(0.0, 8.06, 6 898.61, 0.0, 41.7532)
1.0 x
12
= 50.0, x
16
=60.0 , x
22
= 20.0,
x
23
= 60, x
31
= 30.0, x
38
=80.0,
x
41
=50, x
44
= 30, x
47
= 50,
x
54
=30.0, x
55
= 70.0.
p
14
= 6640, p
16
= 50,
p
18
= 30, p
19
= 90,
p
20
= 410.
(0, 0, 6 640, 0, 40.4)


Fig. 6 Transportation cost different precision levels
6 Analysis of the Solution

p
=0.0 represents the model with unrealistic/impossible
parameters whereas
p
=1.0 is the model with risk free
parameters. Decreasing values of
p
indicates that the model
tends to change from risk free parameter model to
unrealistic/impossible parameter model.
In Table 2, the achievement vector represents the
achievement of the each goal at different precision levels. It
may be seen that the first goal is fully achieved i.e., there is no
violation in the availability at each precision level. The second
goal, i.e., demand goal is also satisfied to maximum extent at
each precision level. It may be noted that at
p
=1, the demand
at each destination is fulfilled (as in this case the total
availability is the same as the total demand) and at
p
=0, the
deviation in the total demand is 0.8 (which may also be seen
from data table considering the upper bounds). From Table 2,
we also find that the transportation cost objective function has
computed to its possible minimum value at each precision
level. At impossible/most risky parameter, transportation cost
is calculated as 11 399.24 i.e. at the precision level
p
=0 and
value 6 640.00, if we use risk free parameter i.e. at the
precision level
p
=1. From the graph (Fig. 6), one can find
how the transportation cost changes as a function of (i.e.,
p
).
For instance, from the graph, the transportation cost
corresponding to fifty-fifty possibility is about 8 219.1.
Brief analysis of the result at a particular precision level say

p
=0.6 is presented below. Different values of x
ij
represent the
amount of coal to be transported from C
i
th colliery to W
j
th
washery and different values of n
i
and p
i
represent the
1.0
T
r
a
n
s
p
o
r
t
a
t
i
o
n

c
o
s
t

(
R
s
.
)

CHAKRABORTY Ananya et al. / J Transpn Sys Eng & IT, 2010, 10(6), 5363

deviations from respective goals. Each value in the
achievement vector given in the table indicates the total
deviation from respective goals. On analysis of the solution
one can infer how far the goals have been achieved at the
precision level
p
=0.6. From the result it is seen that the
deviation variables for the first goal n
i
and p
i
, i=1, 2, , 5 are
all equal to zero which shows that all the materials are shipped
out from all the collieries. Thus the first goal is fully achieved.
In the second goal, management desires to fulfill the demand
of the washeries but there is a total deviation of 24.4 tons from
the total demand at the precision level
p
=0.6. We got n
9
=18.4
and n
11
=6.0 which shows that the demand at washeries W
4
and
W
6
is not be fulfilled with the shortage of 18.4 and 6.0 units
from their respective demands. Thus goal 2 is partially
achieved. The deviation p
14
=7 845.647 in the third term in
achievement vector shows the optimal (best compromise)
transportation cost. The fourth goal is fully achieved, as there
is no deviation for this goal. This shows that the transportation
of coal from different collieries to different washeries for
which maximum time is required has been avoided. Similar
interpretation may be given for goal 5 with respect to time
minimization. Thus, the goal for minimization of time for
transportation is also reasonably achieved.
7 Conclusions
In this paper an effort has been made to develop a model to
minimize the transportation cost as well as to minimize the
transportation time. As the parameters in real world problems
generally are not found in precise form, we incorporated
imprecise parameters in the model. In the proposed model we
find that the objective function Transportation cost (11) and
the resulting goal programming model (30) with (25)-(29) is a
function of the membership function. Both the objectives,
minimization of the transportation cost and minimization of
transportation time are reasonably achieved. Also in Fig. 6 we
see how the transportation cost changes with the change in
membership function. Using fuzzy parametric programming
and preemptive goal programming, the solution has been
obtained which is a good compromise solution. The model
takes care to minimize the total transportation duration. As the
time minimization in transportation problem is a special type
of problem to handle, we have utilized the concept of
partitioned set (of time) and the well-known preemptive goal
programming approach. Although in this model, linear and
exponential membership functions have been considered,
different types of membership functions may also be
considered as per suitability. The proposed technique has been
illustrated with a case study. The model may also be used to
solve time-cost trade off transportation problems with several
objectives with real field data.
References
[1] Hammer P L. Time minimizing transportation problem. Nav.
Res. Log. Quart, 1969, 16(3): 345357.
[2] Hammer P L. Time minimizing transportation problem. Nav.
Res. Log. Quart., 1971, 18: 487490.
[3] Garfinkal R S, Rao M R. The bottleneck transportation
problem. Nav. Res. Log. Quart., 1971, 18: 465472.
[4] Szware W. Some remarks on the time transportation problem.
Nav. Res. Log. Quart., 1971, 18: 473485.
[5] Bhatia H L, Swarup K, Puri M C. Time-Cost trade off in a
transportation problem Opsearch. 1976, 13(3-4): 129142.
[6] Ramakrishnan C S. A note on the time minimizing
transportation problem. Opsearch, 1977, 14: 207209.
[7] Sharma J K, Swarup K. Time minimizing transportation
problems. Proceeding of Indian Academy of Science
(Mathematical Science), 1977, 513518.
[8] Sessan C R, Tikekar V G. On Sharma-Swarup algorithm for
time minimizing transportation problems. Proceeding of
Indian Academy of Science (Mathematical Science), 1980,
101102.
[9] Prakash S. Transportation problem with objectives to minimize
total cost and duration of transportation. Opsearch, 1981,
184(4): 235238.
[10] Bhatia H L, Swarup K, Puri M C. A procedure for time
minimization transportation problem. Indian Journal of Pure
Applied Mathematics, 1977, 8(8): 920929.
[11] Shiang T L. The total cost bounds of the transportation
problem with varying demand and supply. Omega, 2003, 31(4):
247251.
[12] Carlsson C, Korhonen P. A parametric approach to fuzzy
linear programming. Fuzzy Sets and Systems, 1986, 20(1):
1730.
[13] Chanas S. The use of parametric programming in fuzzy linear
programming. Fuzzy Sets and Systems, 1983, 11(2): 243251.
[14] Slowinski R A. Multi Criteria Fuzzy Linear Programming
method for water supply systems development planning.
Fuzzy Sets and Systems 1986, 19(3): 217237.
[15] Gupta P, Mehlawat M K. Bector- Chandra type duality in
fuzzy linear programming with exponential membership
functions. Fuzzy Sets and Systems, 2009, 160(22) :32903308.
[16] Zimmermann H J. Fuzzy Set Theory and Its Applications.
Dordrecht: Kluwer- Nijhoff, 1985.
[17] Sakawa M. Interactive computer programs for fuzzy linear
programming with multiple objectives. International Journal of
Man- Machine Studies, 1983, 18(1): 489503.
[18] Ignizio J P. A review of goal programming: a tool for multi
objective analysis. Journal of Operations Research Society,
1978, 29(11): 11091119.