See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/262621660

A case study application of linear programming and simulation to mine

planning

Article  in  Journal- South African Institute of Mining and Metallurgy · June 2012

CITATIONS READS

3 2,386

3 authors:

José Adolfo Carvalho Júnior Jair Carlos Koppe

Copelmi Mineração Ltda Universidade Federal do Rio Grande do Sul
5 PUBLICATIONS   5 CITATIONS    127 PUBLICATIONS   472 CITATIONS   

SEE PROFILE SEE PROFILE

João Felipe C. L. Costa

Universidade Federal do Rio Grande do Sul
158 PUBLICATIONS   469 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

New metodologies to improve blasting fragmentation View project

high perfomance computing in geostatistics: parallel processing, spectral simulation methods View project

All content following this page was uploaded by João Felipe C. L. Costa on 17 April 2018.

The user has requested enhancement of the downloaded file.

 A case study application of linear
programming and simulation to mine
planning
by J.A. de Carvalho Junior*, J.C. Koppe†, and J.F.C.L. Costa†
Synopsis
This paper analyses the impact of the uncertainty associated with the
input parameters in a mine planning optimization model. A real
example was considered to aid in the building of a mathematical
model that represents the coal production process with reference to
the mining, processing, and marketing of coal. This model was
optimized using the linear programming concept whereby the best
solution was perturbed by the stochastic behaviuor of one of the main
parameters involved in the production process. The analysis of the
results obtained permitted an evaluation of the risk associated with
the best solution due to the uncertainty in the input parameters.
Keywords
mine planning, optimization, risk analysis.
Introduction
Although coal accounts for approximately 50
per cent of the national reserve of fossil fuel,
its use in Brazilian energy generation has
fallen continuously in the last few years, and
coal now contributes less than 1.5 per cent to
the total consumption of fossil fuels. This has
prompted mining companies to diversify their
customer base by supplying thermal coal to
other industries such as the petrochemical,
ceramic, paper, and cellulose industries, where
coal is used for the generation of heat, steam,
and/or electrical power for industrial plants;
the food industry, where coal is used in the
process of drying food grains; and the cement
industry, where coal is used in the clinker mix
in the kilns.
This diversification has forced the mining
companies to develop new products to meet
the specific needs of each customer.
Companies that produced thermal coal with a
calorific value ranging from 3 000 to 4 500
kcal/kg in the past, with practically no
restrictions on the sulphur content, have had
to increase their range of products. Nowadays,
the local market for industrial-purpose coal
demands products with higher calorific values
(4 700 to 6 800 kcal/kg) and a lower sulphur
content (usually below 1.2 per cent). In this
context, and with the constant expansion of
The Journal of The Southern African Institute of Mining and Metallurgy
J
o
u
r
n
a
l
P
the global economy, local coal-mining a
companies are continually looking for p
alternatives that facilitate the production of
better products with competitive prices in e
comparison with other energy-yielding r
products such as natural gas, fuel oil, and
imported coal. An additional complexity is
added to production planning by the fact that
major coal producers frequently operate
multiple mines with multiple pits feeding
multiple processing plants within these mines.
The alternatives commonly chosen to
attend to multiple customers in this multiple-
option production scheme include the
following:
(i) Simultaneously mining multiple
mines, multiple pits, and multiple
benches (seams) to increase the
availability of run-of-mine (ROM)
coal with different characteristics
(ii) Washing of coal from each seam
separately to enhance the yield at the
plant and approximate the character-
istics of the washed coal to the
desired final product
(iii Minimizing the mining dilution so
that the characteristics of the ROM
coal are as close as possible to the
final saleable product.
The definition of the production strategy to
meet a certain market demand is usually
achieved by testing a few production
scenarios: the finally chosen method should
generate the highest economic benefit.
However, this methodology does not guarantee
that the chosen solution is really the best
possible one. With the use of linear
programming, it is possible to find the best
production strategy to meet the demand of
each market.
* CopelmiMineração LTDA, Porto Alegre, Brazil.
† Mining Engineering Department, UFRGS,Porto
Alegre, Brazil.
© The Southern African Institute of Mining and
Metallurgy, 2012. SA ISSN 0038–223X/3.00 +
0.00. Paper received Aug. 2010; revised paper
received Dec. 2011.
L
VOLUME 112 JUNE 2012 477
A case study application of linear programming and simulation to mine planning
This problem can be tackled from an operational research
(OR) perspective. An adequate thermal coal can be produced
with a minimum cost and/or with a higher possible profit.
The production depends on a variety of parameters and
restrictions such as raw material availability, coal quality
(different coal seams), specific product requests, recovery in
the processing plant (which depends on the raw material
used and on the characteristics of the desired product), and
the mining and processing costs. These parameters specify
the production of coal within certain limits.
There are various models and mathematical programming
tools for solving this kind of problem. Among the several
optimization techniques developed, the linear programming
(LP) method has been the most studied, and hence the most
widely used in several of the applied sciences. LP is not only
the simplest technique in mathematical programming, but
also the most versatile, offering the possibility of analysing
the appropriateness of the chosen model concurrent with a
guarantee for obtaining its global best, if it exists.
Due to their complex behaviour, some of the process
parameters, can be regarded as stochastic. To quantify the
impact of these parameters on the optimized outputs derived
from LP, simulation techniques can be used, taking into
consideration not only the averages or the deterministic
patterns, but also the statistical distributions of the
parameters considered.
This paper introduces a methodology for evaluating the
impact of the uncertainty associated with input parameters on
a given objective function using an optimization program that
is based on LP. A case study concerning a major Brazilian
coal-producing company illustrates the process by which the
economil benefits of using the optimal solution are evaluated
in comparison with the solution conventionally used by this
particular mining company. Moreover, incorporating the
uncertainty associated with the input parameters permits an
assessment of the risk in reaching or not reaching the
optimized solution.
Model of linear programming
According to Prado1, LP is a technique that allows the
merging of several variables on the basis of a linear function
of effectiveness (objective function) while simultaneously
satisfying a group of linear restrictions for these variables.
The construction of a LP model follows three basic steps
(Ravindran et al.2):
(i) Identifying the unknown variables to be determined
and representing them using algebraic symbols
(ii) Listing all the restrictions of the problem and
expressing them as linear equations in terms of the
decision variables defined in the previous step
(iii) Identifying the objective or criterion for optimizing
the problem and representing it as a linear function
of the decision variables. The objective can be either
a maximizing or a minimizing function.
Given an LP problem consisting of many variables in the
maximization or minimization mode from a specific linear
function, which is also known as the objective function (OF),
the variables are submitted to a group of restrictions that are
also linear. Generally, a problem in LP is formulated as
follows. The restriction of ’no negativity in the decision
variables’ constitutes a necessary condition for the
L
478 JUNE 2012 VOLUME 112
application of the solution algorithm. Although this
restriction usually holds good due to the nature of the
variables used in the model, under certain circumstances it
can lead to situations where the variables become
unrestricted. In this situation, an artificial variable should be
used to substitute each unrestricted variable by the difference
between the other two variables for which the restriction of
’no negativity’ is applied (Loesch and Hein3).
Briefly, the algorithm can be described as follows:
(i) Maximize or minimize the OF of the type
or
[1]
[2]
(ii) Submit the variables to the following restrictions:
[3]
Where xn are the decision variables, cn are parameters of
the OF, amn are parameters of the restriction equations, bnn
is an independent term of the restriction equations, n is the
number of decision variables, and m is the number of
functional restrictions.
The values of all the coefficients, or control parameters,
are known during the modelling of the problem. These coeffi-
cients can have either deterministic or probabilistic character-
istics depending on the nature of the problem modelled
(Montevevechi4).
Algorithms for the solution
The simplex method is the algorithm typically employed in
the solution of linear programming problems (LPPs). This
technique uses the available tool in linear algebra to
determine the best solution for the LPP by an algebraic
iterative method.
Generally, the algorithm is created by a viable solution of
the system of equations that constitute the restrictions of
LPPs, and starting from that initial solution, it identifies
either new viable solutions for the same or a better value for
the problem than the current one. The algorithm, therefore,
possesses (i) a choice criterion that provides the opportunity
for always finding new and better vertexes of the convex
boundary of the problem, and (ii) a stop criterion that can
verify whether the chosen vertex is the one giving the
optimum solution (Goldbarg5).
The basis for the simplex algorithm is in the formatting of
the limited area for the restrictions, common in all OR
problems (Dantzig6). Such an area is called simplex. Any two
points selected in the outline of a feasible region, when
united by a line, result in a line positioned entirely inside the
simplex. Starting from the verification, the search for the
solution in OR problems becomes limited to the extreme
points of the simplex area. This observation has facilitated
the development of an algorithm of low computational
complexity for solving this problem by Dantzig6.
The Journal of The Southern African Institute of Mining and Metallurgy
A case study application of linear programming and simulation to mine planning
J
The algorithm of LP can be explained using a simplified the process parameters that are not deterministic. This group o
graphical sketch; however, the illustration is limited to two of answers can be associated with the uncertainty space and
decision variables. The example used to illustrate the method can be used to quantify the inherent risk of the given
u
uses the following OF: process. r
[4] According to the functional visualization proposed by n
Menner11, in simulation studies, the models possess (i) a a
The OF is subjected to the following restrictions: number of entrances or inputs x1, x2,..., xr; (ii) a number of
l

[5]
70
P
a
60
After plotting the lines corresponding to the above p
restrictions, an area within which all restrictions are simulta- Restriction
e
neously satisfied (hatched region in Figure 1) is defined. The
polygon defined by the corners A, B, C, D, and E is known as
50 r
Restriction
the ’admissible region’. A
To define the optimum solution, the line corresponding to 40 B
the OF, for a specific value of Z, is initially plotted. In the
next step, various values are defined for Z, and the OFs for X2
Restriction
each defined Z are plotted. A group of parallel lines, known 30
as the lines of the OF levels, is obtained (traced red lines in
Figure 2). The OF grows towards the direction of its gradient
(arrow in Figure 2), and the optimum solution is obtained at Feasible
C
20
the maximum OF that intersects the feasible region. Region
In a two-dimensional solution, the feasible region is
shown in plan. The equation that represents the OF can be 10
depicted by a vector, so that it is possible to reach the highest
point by following the direction of improvement of the OF as
D
determined by the vector inside the space of feasible 0
solutions. The search guarantees that (i) one of the extreme 0 10 20 30 40 50
points respectively maximizes or minimizes the OF, and its X1

value is the maximum or the minimum, respectively; (ii) once

Figure 1— Graphical representation of the algorithm of LP as derived
the search for the highest point is restricted to the space of
from the restrictions presented in Equation [5]
viable solutions of the problem, the optimum vertex satisfies
the group of restrictions composing the problem of LP
(Fogliatto7). 70
This reasoning can be extended to problems of larger
dimensions. Bazaraa8 establishes that while searching for the
highest point during the tracking of the space of viable 60
solutions for an LP problem, the solution should correspond
to one of the extreme points of the simplex. In practice, this
can correspond to a point, a straight line, a plan, or any other 50
pattern with a larger dimension.
The detailed mathematical presentation of the simplex
algorithm can be found along with practical examples in most 40
of the articles referring to LP; see, for example, Ravindran X2
et al.2, Loesch and Hein3, Goldbarg5, Dantzig6, Bazaraa8, and
Bronson9. 30

Stochastic simulation
20
While creating a representative model of any industrial
process in which the behaviours of all the parameters are not
perfectly known, it is not possible to be certain about the 10
results generated by the model. However, it is always
possible to use methodologies of risk analysis to mitigate the
effects of the uncertainty and to visualize the risk problem 0
(Motta10). 0 10 20 30 40 50
X1
For each group of restrictions that encircle the problem, a
group of answers can be obtained from the model using Figure 2— Graphical explanation for the optimum solution (point):
simulation techniques to imitate the stochastic behavior of parallel traced lines represent the lines of the several OF levels
L

The Journal of The Southern African Institute of Mining and Metallurgy VOLUME 112 JUNE 2012 479
A case study application of linear programming and simulation to mine planning
parameters related to the system p1, p2,..., pt; and (iii) a
number of exits or results y1, y2,..., ys, which are the results
defined by a function of the type f (x, y).
Simulation in the sense used in this study constitutes
some mathematical techniques used to mimic any type of real
operation or process (Freitas12). There are a few other
definitions for the term ’simulation’. According to Schriber13,
simulation in modelling implies a process or system that
imitates the responses of the real system during a succession
of events occurring over an extended period.
Pegden et al.14 present a more comprehensive definition
of the simulation process. According to these authors, a
simulation should fulfill the following objectives:
(i) Describe the behaviour of a certain system
(ii) Help in building a theoretical framework and explain
an arbitrary hypothesis, given the observations
provided by the simulation model
(iii) The model finally adopted should be able to predict
new responses due to either a modified system or
modifications in the production processes.
Contrary to the optimization methods, simulation does
not yield an optimal result. Simulations provide a set of
responses for a certain group of variables that condition the
system (Figure 3).
The main advantages of a simulation process include the
following (Pegden et al.14; Banks and Carson15):
(i) Simulation is simpler to use than an analytical
solution
(ii) Although analytical optimization models normally
require many simplifications to obtain a treatable
solution, simulation models do not require these
simplifications
(iii) Any hypothesis about how certain variables
influence the system can be evaluated using the
simulation model
(iv) An appraisal of the variables that are more relevant
to the performance of the system and details of the
interaction between the input variables leading to a
certain response becomes possible
(v) New variations in the production process can be
screened before practical implementation and their
performances can be assessed.
Some systems do not show any probabilistic variables
and hence are known as deterministic. In these circum-
stances, the responses are conditioned only by the input
variables and their interrelations. In various situations, the
Input Variables
Exit Variables
x1
x2
x3
Figure 3—Graphical representation of a simulation process
L
480 JUNE 2012 VOLUME 112
system is controlled by random input variables and,
consequently, a stochastic simulation is used to obtain the
response of the system to a set of inputs. In this latter
situation, the responses are also variable and are modelled
using a probability function.
In the following section, a simulation model is presented
that mimics the production process of thermal coal. The
model is constituted by the following factors: (i) the input
variables, x: the amount of ROM coal originating from
different seams or mines; (ii) the parameters of the system, p:
the operational cost associated with the mining and
processing of the ROM, the indexes of recovery in the
improvement process, and the selling prices of the different
products; and (iii) the exit variables, y: the cash flow and the
values of the liquid assets available. In these terms, the two
groups that contain the input variables and the parameters of
the system comprise the elements associated with
uncertainty, and these can therefore be treated as random
variables.
Settings, products, and production routes
Similar to other mineral commodities, coal production can be
divided into two main stages: exploitation and processing.
Coal mining occurs by open-pit mining, whereby the topsoil
and other sedimentary formations that cover the coal seams,
constituting the overburden, are stripped first, uncovering the
coal seams underneath; these are later mined in an individual
and selective manner (Hartman16). Generically, the mining
operation involves the production of large amounts of waste
material for each ton of coal extracted, constituting one of the
more costly operational stages of the productive process.
The case under consideration is at Copelmi Mineração
Ltda. Copelmi is the largest private coal-mining company in
Brazil, supplying 80 per cent of the industrial market and 18
per cent of the total Brazilian coal market. Its facilities are
located at Rio Grande do Sul State, in the southernmost part
of the country.
The company currently operates four coal mines,
commercializing approximately 1.5 Mt of coal per year. Its
flexibility in producing at multiple mines enables a diverse
product range with calorific value products ranging from
3100 kcal/kg up to 6000 kcal/kg.
Copelmi’s main products are steam coal for a power plant,
a petrochemical plant, and a large paper mill, all located
within a maximum distance of 150 km from the mines.
During the period analysed, four mines were under
simultaneous operation: Recreio Mine, ButiáLeste Mine,
Faxinal Mine, and the Cerro Mine. Each mine exploited a
number of coal seams with different qualities and reserves.
After the mining stage, the coal from the mines is
transported to a washing plant, where it is processed to
obtain a saleable product. The coal from each mine is
processed at the closest washing plant. During this
processing stage, impurities such as pyrite and clay present
in the ROM coal are removed to reduce the emission of ash
and gas during coal combustion. When the mine production
exceeds the feed capacity of the plant due to any operational
contingency, the excess ROM is diverted to the stockpile for
future use. The washing process yields several types of
products. The different production strategies are explained in
the specifications of the OF section of the paper.
The Journal of The Southern African Institute of Mining and Metallurgy
A case study application of linear programming and simulation to mine planning
One of the most relevant operational parameters in the
washing process is the yield or recovery. This parameter
corresponds to the ratio between the amount of product
generated and the amount of ROM coal fed into the plant,
which may be stated as follows:
[6]
Where R is the yield (%), Product is amount of product
generated (t), and ROM is amount of run-of-mine coal fed
into the washing plant (t).
In addition to the product, the washing plant generates
tailings, which are disposed at the tailings dam. These
tailings are predominantly composed of clay and pyrite;
however, due to the inefficiency of the processing plant,
organic matter can be added to the tailings. In the latter
instance, depending on the ash content of the tailings, the
tailings can be redirected to be blended with other products.
Mathematical modelling of the production process
The central subject of the mathematical modelling is the
description of the production process and its decision
variables. The quantity (in tons) to be extracted and
processed from each seam can be determined, thus fulfilling
the demand for each product commercialized by the
company. Optimization is achieved when the production
process generates the largest possible operational contri-
bution after abiding by all the restrictions imposed on the
production process, in addition to considering the demands of
the consumer market.
The decision variables are defined as follows:
® Qij: amount of ROM coal of j seam extracted from mine
i
® Eji: amount of ROM coal of j seam used (or added)
from the stock of mine i
® Ajilk: amount of ROM coal of j seam extracted from
mine i and processed by the washing plant to obtain a
product k.
The process parameters are, among others: coal yield
(plant recovery), production costs, selling prices of the
products, and the tonnage and quality of the product
specified by the consumers.
Objective function
Once the decision variables are defined, the next step is the
definition of the OF. This function is represented by Equation
[7], and its goal is to maximize the operational contribution
satisfying a certain market demand.
[7]
The operational contribution is the financial result
obtained by the commercialization of the generated products
(gross earnings), except for the costs exclusively associated
with the production process (production costs) (Equation[7]).
This function can be better understood if it is subdivided
into four corresponding portions: mining costs (MC), costs of
stocks (SC), washing costs (WC), and gross earnings (GE).
For this variant, a series of parameters representing the
economical, operational, commercial, and geological factors
are used, which can present either deterministic or
The Journal of The Southern African Institute of Mining and Metallurgy
J
probabilistic behaviours depending on the methodology used. o
The four portions that constitute the OF can be expressed in
the form of Equations [8] to [12].
u
r
[8] n
a
[9] l
[10]
P
a
[11] p
e
r
[12]
Where: i = index of the mines; j = index of the seams; l =
index of the improvement plants; k = index of the generated
products; REMi = stripping ratio of mine i (m³/t ROM); CEi
= cost of waste removal from mine i (R$/m³); CDi= blasting
cost of mine i (R$/t ROM); CCi = cost of extraction and
transport of coal from mine i (R$/t ROM); Qji = amount of
coal from seam j extracted from mine i (R$/t ROM); CUEi =
cost of handling unit stock of ROM coal in mine i (R$/t
ROM), Eij = amount of ROM coal of seam j moved on the
basis of the stock of the mine i (t ROM); CBlk = processing
cost of ROM coal in the washing plant for the production of
substance k (R$/t ROM); Ajilk = amount of seam j of mine i
fed and processed in the washing plant to obtain product k
(%); Rjilk = yield from seam j of mine i processed in washing
plant to obtain product k(%); and PVk = selling price of the
product k (R$/t).
Restrictions
The restrictions defined in the outline conditions form the
system and define the domain of the possible values for the
decision variables. In the current case study, the group of
restrictions comprises the following:
® Geological proportion of the seams (percentage of seam
j in the total quantity mined) in each mine
® Maximum capacity of ROM coal production from the
mines
® Maximum coal tonnage able to be fed into each
washing plant
® Recovery of each product at each washing plant
® Market demand
® Qualitative/quantitative restrictions.
Table I shows typical numeric values for the restrictions
in one of the scenarios examined in the case study.
This group of restrictions defines a mathematical
formulation formed by 64 linear equations represented in
terms of decision variables. The theoretical background and
the mathematical formulation used to represent the group of
restrictions in this case study are found in Carvalho17.
An optimization model composed of both the OF and the
group of restrictions after considering the simplifications
detailed above was implemented using Excel®, using the
educational version of the program ’What's Best® 7.0’ as an
optimizer (Lindo Systems18).
L
VOLUME 112 JUNE 2012 481
A case study application of linear programming and simulation to mine planning

Table I

Typical numeric values for the restrictions in one of the scenarios examined in the case study
Maximum capacity of ROM coal production from mines
Mines
Recreio Faxinal Cerro Butia Leste

Capacity of ROM Production (t/month) 60 000 75 000 50 000 50 000

Geological proportion of the seam in each mine

Mines
Seam proportion (%) Recreio Faxinal Cerro Butia Leste

AB Seam 16% 0% 0% 0%
S Seam 34% 39% 81% 50%
S3 Seam 10% 0% 0% 0%
L Seam 6% 37% 0% 0%
M Seam 24% 0% 19% 50%

Maximum coal tonnage able to be fed into each washing plant

Washing Plants
Recreio Faxinal Cerro

Fed capacity (t/month) 60 000 75 000 50 000

Market demand and qualitative/quantitative restrictions

Products
CV54 CV47 CV35 CV30 CV20

Market demand (t/month) 21 000 5 000 21 000 24 000 3 500

Calorific value (kcal/kg) 3 100 3 700 4 700 5 200 6 000
Maximum ash content (%) 54 47 35 30 20
Maximum sulphur content (%) 1.2 1.2 1.1 1.1 1.1

Results of optimization
The outputs generated by the optimization model have been
analysed, taking into account the different production
scenarios. The production scenarios correspond to the
monthly production strategy adopted by the company in 10
different periods during the years 2004 and 2005.
The results are analysed in two steps: the first quantifies
the earnings that would be obtained by using the
optimization model, and the second verifies the risk
associated with each production scenario, considering that
the parameters corresponding to the yield (coal recovery at
the washing plant) behave stochastically, not determinis-
tically.
Analysis of the responses
To quantify the earnings that an optimization model can
generate in the production process, a comparison of the
financial results obtained from the optimization model with
those obtained in the same period adopting the non-
optimized production scheme was conducted. The production
strategy proposed by the mine planning staff without using
optimization tools was adopted as the benchmark.
The scenarios chosen for comparison correspond to the
production strategy adopted by the company in 10 different
situations during the years 2004 and 2005. These situations,
denominated ’Scenario 1’ to ’Scenario 10’, represent
conditions where the market or the operation caused a
significant change in the baseline conditions that define the
production strategy of the company.
L
The optimization process yields as the final result the OF
and the production strategy that need to be adopted. These
strategies correspond to the amount of coal that should be
extracted from each mine, its destination in terms of the
processing plant, and the seam that provides the highest
value for the specific OF.
To quantify the earnings that the optimization process
can generate in a non-optimized process, the final results of
the OF obtained from the optimization model were compared
with the final results obtained following a non-optimized
practical planning process adopted by the company, subject to
the same conditions. The relative differences in the values of
the OF for each of the 10 tested scenarios are presented in
Figure 4.
Figure 4 portrays the results for 10 production scenarios,
with a minimum increase of 2.2 per cent and maximum of
8.7 per cent in the gross earnings, when the optimized
scenarios are compared with the respective non-optimized
procedures adopted by the company. An important practical
aspect that needs to be pointed out is that this gain in
earnings can be obtained by optimizing the production
process within its operational limits, without the need for any
additional investment in the process.
In addition to measurable gains as described previously,
the use of an optimization model provides greater flexibility
in the decision-making process. The method provides the
resource to analyse the impact of possible changes in the
production process, such as changes in the quantity of
product sold and increase or decrease in the production costs.

482 JUNE 2012 VOLUME 112 The Journal of The Southern African Institute of Mining and Metallurgy
A case study application of linear programming and simulation to mine planning
J
10% Additionally, the coefficient of linear correlation between the o
Relative difference in gross earnings (%)

group of possible outcomes of the model and each group of

9%
u
8% parameters identifies which parameter is the most relevant to
7% the final outcome of the model. r
6% To illustrate the risk-analysis methodology, one of the n
5% optimized scenarios was chosen and the yield of different a
4% washed coals from different coal seams was considered as
3%
the parameter of stochastic behavior. The probability distrib-
l
2%
utions of the yields were constructed using a historical series
1%
of results corresponding to the previous 12 months. The
0%
1 2 3 4 5 6 7 8 9 10 statistical summary for these variables is presented in P
Scenario Table II.
Two thousand draws from these probability distributions
a
Figure 4—Relative differences in gross earnings (y-axis) for each were retrieved and used to calculate the OF. p
production scenario (x-axis) after comparing the optimized versus the The variability obtained in the OF and in the coal yield e
non-optimized processes are presented in Figure 5, which shows the relative variation r
between the results of the OF generated based on the
simulated parameters and the results obtained by the
Risk analysis of the optimization model optimization process considering coal yield as a fixed
parameter. Analysis of the results obtained shows that:
An optimization model as introduced above searches for the
highest result, assuming that the input parameters are ® The distribution of the results has a low asymmetry.
deterministically known. However, when the model is This indicates that, in practice, there is no tendency for
influenced by parameters that are not perfectly known, there
is no guarantee that the optimized results will be observed in
x < = -0.22 x < = 0.22
reality. 100%
In this situation, three questions arise:
® How different can the practical result be in relation to 80%
the optimized one?
Cumulative Frequency

® What is the probability of the highest financial result

60%
being smaller than a certain expected minimum value?
® Which parameter has the most significant impact on
the OF? 40%

To answer these questions, the yield parameter was

stochastically varied, by which the deterministic values were 20%

replaced by simulated values. These values were randomly

derived based on the probability-distribution function 0%
constructed from a historical series of results (Bratley -50% - 50% -50% -50% -50% -50% -50% -50% -50% -50%

et al.19). Variation related to the optimum value

For each group of simulated values, the OF generates a

Figure 5—Cumulative distribution function for the objective function
different outcome. A statistical analysis of the group of considering the variability in the coal yield. The results are expressed
possible answers shows the space of uncertainty (risk) as the percentage variation in relation to the optimum value obtained
associated with the expected optimal financial result. when the average yield value was used as the input

Table II

Statistical analysis of yield values

Statistical analysis of historical data for the yields
Mine Seam Product Minimum Maximum Average Median Stand. Dev. #Samples

BL M CV30 34.37 50.38 44.14 44.97 4.38 22

BL M CV35 35.67 64.33 53.42 55.70 7.69 26
BL S CV20 12.82 25.64 18.00 18.04 2.51 67
BL S CV30 37.05 59.83 51.14 51.83 4.86 158
CE M CV35 26.01 70.27 48.38 47.22 10.14 60
CE S CV35 36.99 76.97 59.50 61.10 8.72 40
FX l CV30 18.84 51.16 39.00 39.00 6.80 132
FX l CV47 61.83 96.07 80.00 83.04 11.19 21
FX M CV30 15.70 42.87 31.24 31.70 6.83 159
FX S CV30 17.27 48.22 29.00 28.55 4.87 236
FX S CV47 38.61 92.76 58.71 57.37 10.88 95
CV30 CV20 20.36 33.18 28.00 27.96 2.51 67
REJ CV54 12.10 45.42 25.00 24.09 5.90 228
L

The Journal of The Southern African Institute of Mining and Metallurgy VOLUME 112 JUNE 2012 483
 A case study application of linear programming and simulation to mine planning
over- or underestimation of the OF value in relation to
the optimum value obtained when the average yield
was used as the input
® The results present a high variability, ranging from −42
per cent to +39 per cent in relation to the optimum
value of the OF, and 90 per cent of the values are
between −22 per cent and +22 per cent.
To quantify the influence of each stochastic parameter in
the results of the OF, the coefficient of linear correlation was
calculated for the simulated values with respect to each
parameter, and, subsequently, the OF values were calculated.
The results of these calculations are shown in Figure 6.
From Figure 6, it is noted that all the simulated
parameters show a positive correlation with the results of the
OF.
Discussion and conclusions
The results obtained fom mathematical modelling tools,
optimization using LP, and risk analysis, suggest that
financial gains can be obtained by using these techniques in
mine planning. With these tools, the decisionmaker can
evaluate a series of choices and hypotheses regarding the
production process and thereafter conveniently propose
structural changes in the process to generate a higher profit.
Among the 10 scenarios analysed, the results of the
optimization process indicate that the economic gains could
increase by an average of 5.8 per cent, compared with the
results obtained by the production strategies originally
adopted by the company without any optimization process.
The results also show a large variability in the OF when
the yields for the various coal products are used as stochastic
inputs. This variability in the input is probably overestimated
due to the use of historical data in the simulation process.
The historical data used corresponds to a period of 12
months, during which mining occurred in different locations
within the deposit. This fact can lead to significant variations
Rec. Rej-CV54
Rec. M-FX-CV30
Rec. S-BL-CV30
Rec. S-FX-CV47
Recovering indexes
Rec. S-CE-CV35
Rec. l-FX-CV30
Rec. M-CE-CV35
Rec. M-BL-CV35
Rec. S-FX-CV30
Rec. M-BL-CV30
Rec. CV20
Rec. l-FX-CV47
Rec. -Sub-CV35
0 0.1 0.2 0.3 0.4 0.5
Correlation coefficient
Figure 6—Hierarchy of the linear correlation coefficients (x) between
the results of the objective function and the simulated coal yield for
various saleable coals (y)
L
484 JUNE 2012 VOLUME 112
View publication stats
in the product quality and yield. Hence, geostatistical
simulation should be used to build coal-yield models and,
after these simulations of the models are completed, the
results obtained should be used to feed the yield values into
the optimization process.
The use of a representative mathematical model for the
production process generates other benefits, in addition to the
tangible gains described previously. It allows a compre-
hensive understanding, transparency, and interaction of the
production process, helping to quickly determine the
influence of various factors on the response by different
parameters related to this process. Consequently, the model
can be used not only for the optimization of the current
production process, but also to test alternatives such as the
evaluation of new mine areas or the commercialization of
other products. The correlation coefficient relating each
stochastic parameter and the answers obtained by the OF can
be used to rank the parameters that generate the greatest
impact on the final value of this function. After the definition
of the most relevant parameters, their behaviour can be more
closely followed, and whenever possible, be directed toward
values that improve the result of the OF.
The use of simulation techniques provides the
opportunity to quantify the impact that the stochastic
parameters produce in the optimum response. The stochastic
approach enlarges the range of information available to the
decisionmaker, allowing him or her to evaluate the risk of a
given result, instead of depending on a single deterministic
value of the OF.
References
1. PRADO, D.S. Programação Linear. Belo Horizonte, MG, Desenvolvimento
Gerencial, 1999.
2. RAVINDRAN, A., PHILIPS, D.T. and SOLBERG, J.J. Operations Research
Principles and Practice. 2nd edn. John Wiley & Sons, New York, 1987.
3. LOESCH, C. and HEIN, N. Pesquisa Operacional. FURB, Blumenau, 1999.
4. MONTEVEVECHI, J.A.B. Pesquisa operacional (Programação linear). MSc.
thesis, Escola Federal de Engenharia de Itajubá, 2000.
5. GOLDBARG, M.C. Otimização combinatória e programação linear: modelos e
algoritmos. Rio de Janeiro, Campus, 2000.
6. DANTZIG, G.B. Linear Programming and Extensions. Princeton University
Press, New Jersey, 1963.
7. FOGLIATTO, F. Pesquisa operacional I – Modelos determinísticos. MSc.thesis,
Universidade Federal do Rio Grande do Sul, Porto Alegre, 2004.
8. BAZARAA, J. Linear Programming and Network Flows. 4th edn. Hoboken,
New Jersey, John Wiley & Sons, 1999.
9. BRONSON, R. Pesquisa Operacional. São Paulo. MacGraw-Hill, 1985.
10. MOTTA, R.R. Análise de investimentos – Tomada de decisão em projetos
industriais. São Paulo, Atlas, 2002.
11. MENNER, W.A. Introduction to modeling and simulation. JohnsHopkins
APL Technical Digest, vol. 16, no. 1, 1995.
12 FREITAS, P.J. Introdução à Modelagem e Simulação de Sistemas. Visual
Books, Florianópolis, 2001.
13. SCHRIBER, T.J. Simulation using GPSS. New York, Wiley, 1974.
14. PEGDEN, C.D., SHANNON, R.E., and SADOWSKI, R.P. Introduction to simulation
using SIMAN. New York, McGraw-Hill, 1990.
15. BANKS, J. and CARSON, J.S. Discrete-event System Simulation. Prentice Hall
College Div, New Jersey, 1984.
16. HARTMAN, H.L. SME Mining Engineering Handbook., SME, Littleton
Colorado, ,1992.
17. CARVALHO, J.A. Jr. Abordagem probabilistica em um modelo de
0.6 programação linear aplicado ao planejamento mineiro. MSc. Thesis,
Universidade Federal do Rio Grande do Sul, 2006.
18. LINDO SYSTEMS INC. Premier optimization modeling tools.
http://www.lindo.com Accessed Jan. 2005.
19. BRATLEY, P.,FOX, B.L., and SCHRAGE, L.E. A Guide to Simulation. Springer-
Verlag, New York, 1987. N
The Journal of The Southern African Institute of Mining and Metallurgy
American Journal of Biological and Environmental Statistics
2015; 1(2): 51-57
http://www.sciencepublishinggroup.com/j/ajbes
doi: 10.11648/j.ajbes.20150102.12
ISSN: 2471-9765 (Print); ISSN: 2471-979X (Online)

Case Report
Profit Optimization Using Linear Programming Model:
A Case Study of Ethiopian Chemical Company
Vishwa Nath Maurya1, Ram Bilas Misra2, Peter K Anderson3, Kamlesh Kumar Shukla4
1
Department of Applied Mathematics and Statistics, School of Science & Technology, The University of Fiji, Lautoka, Fiji
2
Department of Mathematics & Computing Science, Divine Word University, Madang, Papua New Guinea
3
Dept. of Information Systems, and Dept. of Mathematics & Computing Science, Divine Word University, Madang, Papua New Guinea
4
Department of Management, Adama Science and Technology University, Adama, Ethiopia

Email address:
prof.drvnmaurya@gmail.com (V. N. Maurya)
To cite this article:
Vishwa Nath Maurya, Ram Bilas Misra, Peter K Anderson, Kamlesh Kumar Shukla. Profit Optimization Using Linear Programming Model: A
Case Study of Ethiopian Chemical Company. American Journal of Biological and Environmental Statistics. Vol. 1, No. 2, 2015, pp. 51-57.
doi: 10.11648/j.ajbes.20150102.12

Received: October 12, 2015; Accepted: April 15, 2016; Published: June 16, 2016

Abstract: This paper aims for profit optimization of an Ethiopian chemical company located in Adama (Ethiopia) using linear
programming model. Particularly, our present study brings out clearly the necessity of using quantitative techniques for
utilization in Ethiopian company; a factory situated within Adama about 90 kms. from Addis Ababa (Capital of Ethiopia). The
first step comprises data generation. A questionnaire is prepared and circulated amongst company staff both executive and
technical to determine the production, sales and profit during a few months of 2014. The profits varied considerably owing to
subjective approach. It was established that the decisions are undertaken by experienced people without use of quantitative
people and quantitative method. Whole approach applied here is seemingly subjective. A theoretical perspective undertaken for
the present study is review of various different applications of linear programming. The characteristics of base assumptions of
linear programming and its advantages and disadvantages towards establishing its need for optimization are briefly outlined in
terms of its application to the factory. Survey data is analyzed to determine the style of decision making and the problem is
defined. An objective function is created in terms of decision variables of production, sales and profit over a period of time using
the quantitatively available data of these parameters. A linear programming model for company is developed for profit
optimization. The model equations with adequate restraints taking into account manufacturing limitations are solved using
MS-Excel solver. Finally, some conclusive observations have been drawn and recommendations have been suggested.

Keywords: Profit optimization, linear programming model, simplex method, manufacturing limitation, service industries

various fronts during World War II. (Here, the word

1. Introduction
Linear Programming (LP) is a problem-solving approach
developed to help managers make decisions. Numerous
applications of linear programming can be found in today’s
competitive business environment Anderson [1]. The term
linear programming was first used by G.B. Dantzig [2] in 1947
to refer to specific problems of optimization which assume
that both constraints and objective function are linear. As with
other branches of Operations Research, the first applications
of LP are found in military planning activities, how to
distribute men, weapons, and supplies efficiently to the
programming means creating a plan that solves a problem; it is
not a reference to computer programming.) Soon after that, LP
came into wide use in industry, with the most fruitful
utilization in the petroleum, petrochemical and food industries
(extensive references can be found in Dantzig [2] and [3].
Linear Programming is not a new modeling technique: it has
been used routinely for over forty years to describe different
productive and economic systems, and also problems in
scheduling and distribution. The mathematics of linear
programming are well established and presented in number of
books ([3],[4], [5], [6]) while computer packages for solving
 American Journal of Biological and Environmental Statistics 2015; 1(2): 51-57
large LP models are well developed and widely available, e.g.
(Dash Assoc., 1993). Linear programming is one specialized
mathematical decision-making aid. It can be applied to many
problems in the real world, not because the world is linear but
because it is a powerful problem-solving technique. Here, the
researches carried out by Kim [7] and Mehdipoor et al. [8] are
also worth mentioning.
Linear programming, or linear optimization, is a
mathematical method to achieve the minimum or maximum
value of a linear function on a convex polyhedron. This convex
polyhedron is, in fact, a graphical representation of some
constraints as inequalities on/off functional variables. To put
simply, one can achieve the best outcome (e.g. maximum profit
or minimum cost) by using linear programming under specific
settings and constraints. While linear programming is mainly
used in management and economics, it can also be utilized for
some engineering problems. Linear programming uses a
mathematical model to describe the problem of concern. Thus,
linear programming involves the planning of activities to obtain
an optimal result, i.e. a result that reaches the specified goal best
(according to the mathematical model) among all feasible
alternatives. Although allocating resources to activities is the
most common type of application, linear programming has
numerous other important applications as well. In fact, any
problem whose mathematical model fits the very general format
for the linear programming model is a linear programming
problem. Furthermore, a remarkably efficient solution
procedure, called the simplex method, is available for solving
linear programming problems of even enormous size [9]. These
are some of the reasons for the tremendous impact of linear
programming in recent decades [5].
As stated above, linear programming was developed as a
mathematical pattern during World War II to plan
expenditures and returns in order to reduce costs to the army
and increase losses to the enemy. The method was kept secret
until 1947. After the war, many industries began using it. The
founders of linear programming are: G.B. Dantzig who
published the simplex method in 1947, John von Neumann
who developed the theory of duality, and Leonid Kantorovich
- the Russian mathematician who applied similar techniques
before Dantzig, and won the Noble Prize in 1957. The
Babcock and Wilcox applied the linear programming to help
plan a major expansion of the company’s Tubular Products
Division (TPD) in Pennsylvania. Owen [10] has also used the
linear programming method to design antenna array patterns
that suppress interference from certain directions.
2. Statement of the Problem
Linear programming is a set of techniques and methods
inferred from mathematics and other sciences which can play
an efficient role in improving the management decisions.
Although it is still regarded as a new science, but it has well
proved to be capable of solving problems such as production
planning, allocating resources, inventory control, and
advertising. Those managers who care about the best
outcomes for their decisions cannot be indifferent to this.
52
Wijeratne and Harris [11] proposed that the linear
programming model is used by the managers to determine the
most economical arrangement of finance, to arrange the best
times to start and finish projects, and to select projects to
minimize the total net present cost of capital. Linear
programming optimizes (maximizing or minimizing) a
dependent variable subject to a set of independent variables in a
linear relationship, given a number of linear constraints of
independent variables. The value of dependent variables which
is the value obtained from solving the problem, is subject to the
independent variables set by the decision maker (or determined
by solving the problem). The dependent variables are usually
set as objective function which may be one of the economic
concepts such as profit, cost, income, production, sales,
distance and time, etc. The independent variables in linear
programming are known as variables of unknown value, and
the decision maker has to calculate the value of such variables
by solving the problem [13]. In Ethiopia, most decisions were
taken by government or non-governmental organizations,
whether it is profitable or not for companies, manufacturing or
service industries are based on trial and error. Qualitative
decisions, like intuition, judgmental approaches are more
dominant and the application of model based decision making
like optimization techniques, e.g. linear programming models
have little or no application. Hence, it is initiated by author to
conduct an assessment of the application of linear programming
in this particular company as a case study.
3. Objectives and Research Questions
3.1. Objectives of the Study
The main objectives of this paper are:
To formulate a linear programming model that would
suggest a viable product-mix to ensure optimum profit
for Company.
To highlight the peculiarities of using linear
programming technique for the Company and prove that
despite obstacles, the application of the technique in
determining the product-mix of the company would be
more profitable than otherwise.
To know about the constraints of the company regarding
cost, resources.
3.2. Research Questions
How is the company currently making decision on the
allocation of resources for the production of its
products to maximize its sales and profit?
Whether the company uses qualitative decisions or
employs mathematical/statistical models or both?
Whether the current method adopted in making decision
is effective or not?
Whether the resources available are sufficient for
production of the products to satisfy the demands of the
customers?
53 Vishwa Nath Maurya et al.: Profit Optimization Using Linear Programming Model:
A Case Study of Ethiopian Chemical Company
4. Significance of the Study
It helps to understand the best way of making decisions
using quantitative models in order to determine its optimal
product-mixes that can maximize its profits subject to the
scarce resources it has. The study will provide a deep
understanding and insight of the applications of linear
programming models in industries and how to apply such
models in practical and real world experience. To other
researchers of similar interest who are willing to undertake
further investigation on the topic, this research document can
be used as a secondary information source.
This research paper is structured in seven sections.§1 is
introduction and it concerns with background of the study and
the company. § 2 – 4 deal with the statement of the problem,
objectives of the study, research question, and significance of
the study. §5 deals with data and methodology. Linear
programming and its applications are discussed in §6.
Presentation of data, analysis and results are also discussed in
§6. Major findings and conclusions are given in§ 7. Finally,
some recommendations are illustrated in the end.
This study was conducted for three reasons. Firstly, there
was no reliable and comprehensive study to examine the role
of linear programming in companies in Ethiopia. Secondly, it
might pave the way forward for the company, policy makers,
and financial institutions to understand the different roles of
institutions in the development process. Finally, this study
advances the knowledge of linear programming in decision
making.
5. Data and Methodology
5.1. Method of Data Collection
In order to gather the relevant data from the company
producing sulfuric acid, oleum and Aluminum sulfate utilizing
partially locally available raw materials. This paper was based
on primary and secondary sources of data. Especially to get
information on the decision making practice of the company
under study, an interview was conducted with the
manufacturing manager and the sales managers of the
company and data collected on the unit costs of production of
the products, the unit selling price of the products and also
contacted the sales and the purchasing, production and
marketing managers of the company. And additionally
secondary data sources were used to get accurate information.
5.2. Method of Data Analysis
Collected data were presented in a narrative form and
analyzed using linear programming model. Since the purpose
of this study was to develop linear programming model for the
collected data from the company, the authors tried to
transform the data into a linear programming model and
solved the model (problem) using simplex algorithm using by
applying MS-Excel solver in order to determine the optimal
combination of the products of the company that can
maximize its profit within the available scarce resources.
Simplex algorithm was preferred over graphical approach
because of this method can help to solve linear programming
problems of any number of decision variables. Excel solver
was preferred for accuracy purpose.
6. LPP Model and its Application
Linear Programming is a mathematical technique for
generating and selecting the optimal or the best solution for a
given objective function. Technically, linear programming
may be formally defined as a method of optimizing (i.e.
maximizing or minimizing) a linear function for a number of
constraints stated in the form of linear inequalities.
According to Fagoyinbo [5], Williams [12] and Wood and
Dantzig [13] the problem of LP stated as that of the
optimization of linear objective function of the following
form:
Z = c1x1 + c2x2 (Objective function),
subject to the linear constraints:
a11 x1+ a12 x2(≤ or≥) b1,a21 x1+ a22 x2(≤ or≥) b2,…… ,
am1 x1+ am2x2(≤ or≥)bm,x1, x2(≤ or≥) 0.
6.1. The Proposed Model
The objective function and constraints were used in the
following general form:
Maximize Z = P1X1 + P2X2 subject to constraints:
C11X1 + C12X2≤B1, C21X1 + C22X2≤B2, C31X1 + C32X2≤B3,
C41X1 + C42X2≤B4, C51X1 + C52X2≤ B5, X1, X2≥ 0,
where:
Z = total profit contribution of Aluminum sulphate and
Sulphuric acid,
Pi = profit contribution coefficients expressing per unit
contribution to the profit equation,
X1, X2 = the set of unknowns to be determined, i.e. the tons
of Aluminum sulphate
And Sulphuric acid produced by company,
C’s = the numerical values that expressed in per unit usage
of the right hand side,
B1, B2,….,B5 = the resource values which were fully
utilized.
Estimates of the variables are presented in tables. The
optimum values of the different brands produced by the firm
shows the combination (product mix) obtained through the
application of linear programming model.
6.2. Assumptions of the Model
For applying the proposed model to case of profit
optimization, following assumptions were taken into
consideration:
(i) Proportionality assumption: The contribution of each
activity to the value of the objective function Z is
proportional to the level of the activity xj, as
 American Journal of Biological and Environmental Statistics 2015; 1(2): 51-57
represented by the terms cjxj in the objective function.
Similarly, the contribution of each activity to the
left-hand side of each functional constraint is
proportional to the level of the activity xj, as
represented by the terms aijxj in the constraints.
(ii) Additive assumption: Every function in a linear
programming model (whether the objectives function
or the function on the left-hand side of a functional
constraint) is the sum of the individual contributions
of the respective activities.
(iii) Divisibility assumption: Decision variables in a linear
programming model are allowed to have any (real)
values, including non-integer values that satisfy the
functional and non-negativity constraints. Since each
decision variable represents the level of some activity,
it is being assumed that the activities can be run at
fractional levels.
(iv) Certainty assumption: The value assigned to each
parameter of a linear programming model is assumed
to be a known constant.
6.3. Application and Analysis
This section deals with the presentation and analysis of the
Table 1. Production, sales and gross profit trends for the years 2006-2012.
Year 2006 2007 2008
Net sales 8,206,627 10,432,627 8,341,567
Production(tons) 2,499.408 1,654.404 2,245.226
Gross profit 1,227,241 3,688,072 3,284,513
(Source: Annual bulletin of company, 2012/13)
According to the Annual Bulletin of the company (2012/13),
it has been described as below: As compared to the design
(production) capacity of the company, the actual production
achievement of aluminum sulphate and sulphuric acid is on
average 21.51% and 19.56% only. Even though there is an
improvement in the actual sales of these items\o some degree
over the past four years, compared to the capacity of the
company, it did not exceed annual average sales of 20%.
Though the reason for such low sales is assumed to be the
decline in domestic market the major problems are:
Some users of the products were importing substitute
products for their purpose. However, considering the current
industry development strategy of the country and the attention
given by the government for industry, the company was
expecting more demand for its products in the near future.
Even though the company states the above factors as causes
for low demand and sales of its products, the last author
identified that the annual demand for aluminum sulphate was
on average 6,000 tons per annum and for sulphuric acid it was
5,000 tons per annum. This indicates that the company was not
fully utilizing the currently available demand for its product.
One of the cause for underutilizing the demand was the poor
decision making approach used by the company. Had it used
Quantitative (model based) decision making especially linear
programming, it might have achieved more than what it did.
According to the information gathered from the company
54
data gathered from the company. For collection of the
necessary data, the last author prepared a guide questions
based on its objectives and research questions. He made an
effort to know whether the company applies any model for
their decision making especially whether linear programming
models are applicable for decision making. If decision is not
based on mathematical model, then researchers tried to
investigate the style of decision making of this company.
Through this research it is identified that the company
manufactures two major products namely: aluminum sulphate
and sulphuric acid.
6.4. The Decision Approach of the Company
According to collected data of the Company authors
investigated that the company has annual based sales records.
In order to determine how much to manufacture for the next
year the company do not use any sound mathematical or
statistical decision making models. Instead it decides the
quantity of its products on a simple trial and error or simple
individual’s judgment. For this reason their annual sales plan
and the actual sales they achieve varies every year. When it
comes to the case of the applicability of linear programming
model for the optimal decision of the product mixes.
2009 2010 2011 2012
19,027,529 30,204,051 39,549,397 50,845,928
2,926.932 4,285.742 4,856.498 7,362.096
4,160,415 8,198,872 13,304,129 14,931,836
using linear programming model (MS-Excel Solver) it is
identified below that the company should produce 20tons of
aluminum sulphate per day (20×300=6,000tons per year) and
6.98tons of sulphuric acid per day (6.98×300=2,094 tons per
year) in order to get a gross profit of Birr107,353.17 per day
(107,353.17×300= 32,205,951 per year).
Figure 1. Year wise grass profit in Birr.
6.5. LPP of the Chemical Company
A linear programming model consists of certain common
components and characteristics. The model components
include decision variables, an objective function, and model
55 Vishwa Nath Maurya et al.: Profit Optimization Using Linear Programming Model:
A Case Study of Ethiopian Chemical Company

constraints, which consist of decision variables and
parameters. Decision variables are mathematical symbols that
represent levels of activity by the firm.
Data gathered from the company into a linear programming
model as follows:
A chemical industry owned by the Ethiopian Government
manufactures two major products, aluminum sulphate and
sulphuric acid each of which passes through three different
processes: reaction, filtration, and evaporation. Aluminum
sulphate is a coagulant used for purification of drinking water,
industrial waste water treatment, paper sizing, dyeing,
pharmaceuticals, soap modification, tanning cellulose , etc. so
it is highly demanded by companies like: water supply and
sewerage authorities, pulp and paper factories, leather
industries, textiles, food and beverages, metals, chip woods,
Table 2. Constraint of machine hours on 3 processes of manufacture.
plywood and others. Sulphuric acid is a chemical used in the
production of aluminum sulphate as a raw material or as
medium of production for oleum, phosphates, car batteries,
pulp & paper, steel picking and textiles (brochures of the
company).
According to data collection, the company was
manufacturing 51.5 tons of sulphuric acid per day
(51.5×300=15,450 tons per year) and only 40 tons of
aluminum sulphate per day (40×300 = 12,000 tons per year).
The major constraint the company states for this is demand
which is at most 6,000tons for aluminum sulphate per annum
and utmost 5,000 tons per annum for sulphuric acid. But in
addition, availability of machine hours on 3 processes of
manufacture was also constraining manufacturing of the
products as given below:

Machine hours on Demand in

Products
Reaction Filtration Evaporation tons/annum
Aluminum sulphate (for 40tons per day) 18 8 4 6,000
Sulphuric acid (produced 51.5 tons daily) 24 24 24 5,000
Total available time (assuming 300 working days per year) 7,200 7,200 7,200

According to the production / sales departments of the

company, the production of Aluminum sulphate costing Birr
5,237.69 per ton is sold for Birr8,998.35 per ton, whereas the
manufacturing of sulphuric acid costing Birr 5,708.87 per ton
is sold for Birr 10,313.45 per ton. Thus, the profit earned per
ton of these items is Birr 3,760.66 and Birr 4,604.58
respectively. The objective is to determine how many tons of
these items should be manufactured per day to maximize the
daily profit of the company operating on average of 300 days
in a year.
6.6. Mathematical Formulation of LPP for the Company
The profit maximization objective of the company is
mathematically expressed as:
Table 3. Number of hours of machine time on three processes and the demand of items.
Zmax = 3,760.66x1 + 4,604.58x2.
Decision variables are mathematical symbols that
represent levels of activity.
Letx1 = the tons of aluminum sulphate manufactured per
day, and
x2 = the tons of sulphuric acid manufactured per day.
The model constraints are also linear relationships of the
decision variables; they represent the restrictions placed on the
firm by the operating environment. Four constraints relate to
the number of hours of machine time available on three
processes (reaction, filtration, and evaporation); and demand
also restricts the number of tons of the items that can be
manufactured as presented below:

Machine hours per ton per day* Produced/day

Products Profit /per ton (in Birr)
Reaction Filtration Evaporation (in ton)
Alu. Sulphate 0.45 0.2 0.1 20 3,760.66
Sulphuric acid 2.15 2.15 2.15 51.5 4,604.58
Available resource 24 24 24

*The machine hours are roughly converted to per day basis for simplicity of model formulation and calculation while company officials suggested that it is
sometimes not logical to make such conversion.

The constraints of the company are expressed as follows: x1≤20 tons (demand for aluminum sulphate per day),

Maximize Zmax = 3,760.66x1 + 4,604.58x2, x2≤51.5 (sulphuric acid produced per day),

subject to: x1, x2≥0 (non-negativity).

0.45x1 + 2.15x2≤24 hrs. (machine hrs. on reaction), 6.7. Solution of the LPP Using MS-Excel Solver
0.2x1 +2.15x2≤24 hrs. (machine hrs. on filtration), To apply the MS-Excel Solver, first we translate the linear
programming model into its standard form. Thus, the above
0.1x1 + 2.15x2≤24 hrs. (machine hrs. on evaporation), model, written in the standard form, becomes.
 American Journal of Biological and Environmental Statistics 2015; 1(2): 51-57 56

Zmax = 3,760.66x1+ 4,604.58x2 + 0s1 +0s2 +0s3 +0s4 +0s5, 0.25 x1 +s1 =s2, 0.35 x1 + s1 =s3, 0.1 x1 + s2= s3.

subject to: There are five equations involving seven unknown

variables. So, their basic solutions can be obtained by
0.45x1 + 2.15x2 +s1 =24 hrs., 0.2x1 +2.15x2 +s2=24hrs., assigning zero values to any (additional) 2unknowns. Setting
0.1x1+ 2.15x2+s3 =24 hrs., s1=s2 = 0 a basic and feasible solution is

x1+s4=20 tons, x2+s5= 51.5 tons, x1= 0, x2=24/2.15 = 11.16, s3 = 0,s4= 20, s5 = 40.34;

x1, x2,s1, s2,…,s5≥0 (non-negativity). that determines Z = 51,400.08. In the following, we workout
all basic solutions and compute the values of Z:
Eliminating x2 from the first three equations there also result
the equations:

Table 4. Solution of the LPP for the Chemical Company by MS-Excel Solver.

x1 x2 s1 s2 s3 s4 s5 Is soln. feasible Z
0 11.16 0 0 0 20 40.34 Yes 51,400.08
20 15/2.15 =6.98 ” 5 7 0 44.52 ” 107,353.17
− 192.72 51.5 ” <0 <0 212.72 0 No
0 11.16 0 0 0 20 40.34 Yes 51,400.08
20 20/2.15 = 9.3 −5 ” 2 0 42.2 No
− 433.2 51.5 <0 ” <0 453.2 0 ”
20 22/2.15 = 10.23 −7 −2 0 0 41.27 ”
− 867.25 51.5 303.54 86.73 ” 887.25 0 ”
20 ” <0 <0 <0 0 ” ”

Thus, the optimal feasible solution is x1= 20 tons of aluminum sulphate per day, x2 = 6.98 tons of sulphuric acid per day, s1=0, s2=5 hrs., s3=7 hrs., s4=0, s5=44.52
tons, yielding the maximum profit Zmax= Birr 107,353.17 per day.

6.8. Result Analysis quantitative model to forecast for their production

Company produces 20tons of aluminum sulphateperday
(20tons/day×300working days = 6,000 tons/annum) and 6.98
tons of sulphurc acid per day (6.98tons/day×300 working days
= 2,094tons/annum) in order to get a maximum daily profit of
Birr107,353.17 per day (Birr107,353.17 per day ×300working
days =Birr32,205,951per annum). In this way, the company is
left with an idle filtration and evaporation times of 5 and 7
hours per day and unutilized demand for sulphuric acid of
44.52tons per day but the demand for aluminum sulphate is
fully utilized.
7. Conclusive Observations
Based on the data gathered and the major findings of the
research, the following conclusions were drawn:
(i) Linear programming model is applied for profit
optimization of the Ethiopian chemical company
located in Adama (Ethiopia) and the maximum profit
Birr 107,353.17 per day for company was found.
(ii) The company was left with an idle filtration and
evaporation time of 5 and 7 hours per day respectively
and un-utilizing demand for sulphuric acid of 44.52
tons per day but the demand for aluminum sulphate
was fully utilized.
(iii) Even though the company has an annual record of its
production and sales volume it was not employing any
volume and sales; rather the company was employing
a simple trial and error and subjective estimation. It
led the company to misunderstand the existing
demand for their products and to produce less than its
capacity and even than the existing local demand.
(iv) The LPP model can be used for other companies for its
profit maximization.
Recommendations
Based on the findings and the conclusions of the study, we
suggest the following recommendations:
(i) For companies to forecast their production capacity and
sales volume different quantitative models like: linear
regression, linear trend, nonlinear regression, nonlinear
trend, etc. available. Company has annual production
and sales records but the company was not employing
any mathematical or statistical models for its
production or sales forecast rather it depends on a
simple trial and error. The company should employ
linear regression, linear trend and/or nonlinear
regression and trend models to forecast its production
capacity and sales.
(ii) On the major constraint that is considered by the
company was demand for its products. In the 2012/13
annual bulletin of the company, it was stated that some
customers of company were importing substitute
57 Vishwa Nath Maurya et al.: Profit Optimization Using Linear Programming Model:
A Case Study of Ethiopian Chemical Company
products. This indicates that there was a quality and
supply problem. Therefore, company should work on
improving the qualities of its products to meet customer
expectations and to search for other customer groups
who can use the products beyond the current target
markets.
(iii) It is clear that model based decision is important for its
accuracy and objectivity. But such decision making
approach was not widely used in practice. Qualitative
decisions like subjective estimation, intuition and trial
and error were commonly used by company. It is eye
opening concern to the policy makers of company to
shift the model based decision making styles in general.
References
[1] Anderson, D.R.; Sweeney, D.J.; Williams, T.A.; Camm, J.D.;
and Kipp, Martin:An Introduction to Management Science:
Quantitative Approaches to Decision Making, Revised 13th ed.,
South-Western Cengage Learning, 2012.
[2] Dantzig G.B., Programming of interdependent activities: II
Mathematical Model, Econometrica, 17 (3), pp. 200–211, 1949.
doi:10.2307/1905523.
[3] Dantzig, G.B.: Compact basis triangularization for the simplex
method, R.L. Graves and P. Wolfe (eds.), Recent Advances in
Mathematical Programming, McGraw-Hill, New York,
pp.125–132, 1963.
[4] Drayer, W. and Seabury, S.: Linear programming - A case
example, strategy & leadership, 3(5), pp.24-26, 1975.
[5] Fagoyinbo, I.S.: Compendious text on quantitative techniques
for professionals, Ilaro, Nigeria, Jombright Productions, 2008.
[6] Frederick, H.S. and Lieberman, J.G.: Introduction to
Operations Research, McGraw-Hill, Operations research -
1214 pages, 2001.
[7] Kim, C.: Parametrizing an activity vector in linear
programming, Operations Research, 19, pp.1632-1646, 1971.
[8] Mehdipoor, E.; Sadr-ol-ashraafi, S.M. and Karbaasi, A.: A
comparison of canonical linear programming techniques,
Meaty Chicken’ Feed Farming with Linear Programming
Models, Scientific-Research Magazine of Agriculture, 12(3),
2006.
[9] Misra, R.B.: Numerical Analysis for solution of ordinary
differential equations, Lambert Academic Publishers,
Saarbrücken (Germany), 2010, ISBN 978-3-8433-8489-6.
[10] Owen, P. and Mason, J.C.: The use of linear programming the
design of antenna pattern with prescribed nulls and other
constraints, compel: The International Journal for Computation
and Mathematics in Electrical and Electronic Engineering, 3(4),
pp.201-215, 1984.
[11] Wijeratne, N. and Harris, F.C.: Capital budgeting using a linear
programming model, International Journal of Operations &
Production Management, 4(2), pp.49-64, 1984.
[12] Williams, N.: Linear and non-linear programming in industry,
Sir Isaac Pitman & Sons, Ltd., London. Garifinkel (1963). A
solution of the Goddard problem, Journal of SIAM Control,
1(3), pp.349–368, 1963.
[13] Wood, M.K. and Dantzig, G.B.: Programming of
interdependent activities: I General Discussion. Econometrica,
17 (3/4): 193–91949. JSTOR1905522.

 

 
RESEARCH ARTICLE

The Optimization Problem of Product Mix

and Linear Programming Applications: Case
Study in the Apparel Industry

Gera Workie Woubante*

Bahir Dar University, Bahir Dar, Ethiopia

*Corresponding Author: Gera Workie Woubante: likeredg@gmail.com

Abstract
Citation:  Woubante  G.  W.  (2017)  
The  Optimization  Problem  of  
Product  Mix  and  Linear   Industrial development strategy is characterized by the efficient
Programming  Applications:  Case   use of resources at every production stage. The analysis and
Study  in  the  Apparel  Industry.  
Open  Science  Journal  2(2).     efficient utilization of resources are made sustainable by
 
 
effective management decision making techniques employed in
th
Received:  6  January,  2017   the industry. A quantitative decision making tool called linear
 
th
Accepted:    20  March,  2017  
programming can be used for the optimization problem of
 
th
product mix. Understanding the concept behind the
Published:  19  June,  2017  
  optimization problem of product mix is essential to the success
Copyright:©  2017  This  is  an   of the industry for meeting customer needs, determining its
open  access  article  under  the  terms  
of  the  Creative  Commons   image, focusing on its core business, and inventory management.
Attribution  License,  which  permits   Apparel manufacturing firms profit mainly depends on the
unrestricted  use,  distribution,  and  
reproduction  in  any  medium,   proper allocation and usage of available production time,
provided  the  original  author  and   material, and labor resources. This paper considers an apparel
source  are  credited.  
  industrial unit in Ethiopia as a case study. The monthly held
Funding:  The  author(s)  received   resources, product volume, and amount of resources used to
no  specific  funding  for  this  work  
  produce each unit of product and profit per unit for each
Competing  Interests:  The  
author  have  declared  that  no  
product have been collected from the company. The data
competing  interests  exists.   gathered was used to estimate the parameters of the linear
programming model. The model was solved using LINGO 16.0
software. The findings of the study show that the profit of the
company can be improved by 59.84%, that is, the total profit of
Birr 465,456 per month can be increased to Birr 777,877.3 per
month by applying linear programming models if customer
orders have to be satisfied. The profit of the company can be
improved by 7.22% if the linear programming formulation does
not need to consider customer orders.

Open  Science  Journal  –  June  2017     1  

 
Open  Science  Journal  
Research Article
 

Keywords: Apparel sector, Linear programming, Lingo, Optimal,

Product mix.

Introduction

Companies in the world, including Ethiopia, have faced problems on

optimization of production inputs. A company’s endurance in a competitive
market closely depends on its ability to produce the highest quality products at
the lowest possible cost (Kumar, 2010). Ezema and Amakom (2012) emphasized
that organizations in the world are challenged by shortages of production inputs
and low capacity utilization that can consequently lead to low production
outputs. Companies have to create a management style to guide their
performances in processing and resource utilization. Apparel manufacturing firm
profit is significantly affected by the cost of resources and resource utilization
(Arefayne and Pal, 2014). Linear programming is an operational research
technique used to allocate optimally production resources for a firm’s best
practices. It is the most widely used tool (Reeb and Leavengood, 1998) to
determine optimal resource utilization. Different products require different
amount of production resources having different costs and revenues at different
stages of production. Thus, the linear programming problem (LPP) technique will
be used to determine the product mix that will maximize the total profit at a
specified time. It is the best method for determining an optimal solution among
alternatives to meet a specified objective function limited by various constraints
and restrictions (Shaheen and Ahmad, 2015). As Reeb and Leavengood (1998)
stated, it is a planning process that allocates resources—labor, materials,
machines, and capital—in the best possible way so that costs are minimized or
profits are maximized. The LPP then becomes a problem of allocating scarce
resources to products in a manner such that profits are at a maximum and/or
costs are at a minimum (Yahya, 2004). However, managers in companies create
gaps in adopting the method to allocate scarce resources among operations and
providing quantitative analysis for each production period due to lack of
awareness.
An industry consists of a number of product lines, which are groups of items
featuring similar characteristics. Together, these product lines comprise the
product mix, which is the total assortment of products the company offers.
Product mix determination is essential to the success of the industry for a
number of important reasons. First, crafting the proper assortment of products
provides the best opportunity of meeting customers' needs. The customers come
to rely on the industry as their primary resource. By carefully monitoring product
mix and seeking feedback from customers, the firm will be able to make
adjustments as the needs of customers change over time. Second, it shapes the
image of the industry and its brand to maintain consistency in the eyes of the
target market. Third, the industry stays focused on its core business. The
industry may be tempted to add more product lines in an effort to reach more
customers. By doing so, it could be in danger of adding products that appeal only
to a fraction of its customer base while alienating the core customers. As a rule,
core customers are the ones who provide about 80 percent of the business, so

Open  Science  Journal  –  June  2017     2  

 
Open  Science  Journal  
Research Article
 

these customers have to be kept in mind when contemplating changes to the

product mix. Finally, the industry probably faces limitations regarding the
number of products it is able to offer. By focusing on providing the optimum
product mix for its customers, it will be able to weed out slow sellers and make
room for more appealing items.
In Ethiopia, the apparel sector is considered as one of the priority areas of
the government‘s industrial development strategy (ETIDI, 2014). However, the
sector has faced many challenges to determine product mix. It is confronted with
inefficient utilization of resources that makes it difficult to ensure the optimal
product mix for maximal profit, which would also fulfill customer needs. As noted
by Samuel (2012), to modernize the apparel sector, performance was insignificant.
Thus, apparel manufacturing companies must adopt operations research
techniques to enhance best resource utilization that would result in optimal
product mix and total profit. Thus, this paper focuses on product mix
determination based on efficient resource utilization for the Ethiopian apparel
sector by considering a garment factory in Ethiopia as a case company. The issue
addressed here was to determine the product mix for optimal profit with
available resources, using the linear programming technique LINGO 16.0, that
the case company should produce for men wear: Polo shirt, basic T-shirt, mock
neck T-shirt, short pants, and singlet. In line with this, the objective of the study
was to suggest linear programming as a decision tool to determine the optimal
product mix for maximum profit with available resources in the apparel sector.

Methodology

The data collection procedure was quantitative in nature and relied on face-
to-face interviews with members of the management and line supervisors in
accordance with existing records and merely amended to finalize the concepts
relevant to the resources held and consumed and the production volume of each
product in the case company. The relevant information on the amount of
resources used per unit of each product during the month is summarized in Table
1.

Table 1: Resources needed per unit of product.

Resource used per unit of products

Products (T-
Overheads

Finishing
Threads

Shirts and
Cutting
(Meter)
(Gram)
Fabrics

Sewing
Labor
(Birr)

(Birr)

(Min)

(Min)

(Min)

Pants)

Polo T-shirts
315 230 11.6 30.6 1.8 22.7 2
Basic T-shirts 200 110 5 19.1 1.1 5.4 1.3
Mock Neck T-
shirts 195 140 6.3 20.1 1.7 10.4 1.9
Singlets 180 100 4.15 16.5 1.1 4.5 1.3

Open  Science  Journal  –  June  2017     3  

 
Open  Science  Journal  
Research Article
 

Short Pants 280 200 7.5 37.5 2.6 20.1 2.6

Source: 2015 Resources used per unit of products of the case apparel company

The ability to use resources (resource utilization) was recorded as the major
constraints in the case apparel manufacturing unit. Seven constraints (fabrics,
thread, labor, overheads, cutting, sewing, and finishing time) and five costumer
orders for T-shirt products have been identified. Out of the resources used by the
case apparel company, the major items held and consumed are shown in Table 2.

Table 2: Average monthly resources held and consumed in quantity/value terms

in Birr.
Resources
Resource Type Measurement Unit Held Value Consumption Value

Fabrics kg 38,665 20,526.871

Threads Meter 26,638,120 132,887

Labor Birr 1,009,008 590,765

Overheads Birr 4,979,414 2,116,525

Cutting h 5,770 2,320.42

Sewing h 44,505 16,705.7

Finishing h 6,195 2,583

The demand and profit earned from each product during the month for the
case apparel company are depicted in Table 3.

Table 3: Demand and profit earned.

Polo T-

Basic T-

Neck T-

Singlets
Shirts

Shirts

Pants
shirts

Short
Mock

Demand 15886 23916 13319 25667 12864

Profit per Unit 4.22 3.62 3.43 3.10 6.75

Model Formulation

In formulating a given decision problem in mathematical form, one should try

to comprehensively understand the problem (i.e., formulate a mental model) by
carefully reading and re-reading the problem statement. While trying to
understand the problem, the decision maker may decide that the model consists
of linear relationships representing a firm’s objectives and resource constraints.
However, the way we approach the problem is the same for a wide variety of
decision making problems, and the size and complexity of the problem may differ.
An LPP model consists of the following parameters:

Open  Science  Journal  –  June  2017     4  

 
Open  Science  Journal  
Research Article
 

Ø Decision variables that are mathematical symbols representing levels of

activity of an operation.
Ø The objective function that is a linear mathematical relationship
describing an objective of the firm, in terms of decision variables, that is
to be maximized or minimized.
Ø Constraints that are restrictions placed on the firm by the operating
environment situated in linear relationships with the decision variables.
Ø Parameters/cost coefficients that are numerical coefficients and constants
used in the objective function and constraint equations.

Basic Assumptions (Properties) of LPP

Technically, there are five additional requirements of an LPP:

1. We assume that conditions of certainty exist; that is, the numbers in the
objective and constraints are known with certainty and do not change
during the period being studied.
2. We also assume that proportionality exists in the objective and
constraints. This means that if production of 1 unit of a product uses 3 h
of a particular scarce resource, then making 10 units of that product uses
30 h of the resource.
3. The third technical assumption deals with additivity, meaning that the
total of all activities equals the sum of the individual activities.
4. We make the divisibility assumption that solutions need not be in whole
numbers (integers). Instead, they are divisible and may take any
fractional value.
5. Finally, we assume that all answers or variables are nonnegative.
Negative values of physical quantities are impossible; we simply cannot
produce a negative number of textile products.

Linear Programming Model Formulation Steps

In this section, we consider the steps involved in the mathematical

formulation of the problem. LPP is a collection of the objective function, the set
of constraints, and the set of nonnegative constraints.

Step 1: Clearly define the decision variables of the problem, X = (x , x ,…, x ).

1 2 n
Step 2: Write the objective function as a linear combination of the decision
variables, Z = f (X).
Step 3: Formulating the constraints of the problem as a linear combination of the
decision variables.

General Form of the Linear Programming Model

In general, if C = (c , c ,…, c ) is a tuple of real numbers, then the function f of

1 2 n
real variables X = (x , x ,…, x ) defined by
1 2 n
f (X) = c x + c x +…+ c x
1 1 2 2 n n

Open  Science  Journal  –  June  2017     5  

 
Open  Science  Journal  
Research Article
 

is known as a linear function. If g is a linear function and b = (b , b ,…, b ) is a

1 2 n
tuple of real numbers, then g(x) = b is called a linear equation, whereas g(x) (≤ ,
≥) b is called a linear inequality. A linear constraint is one that is either a linear
equation or a linear inequality. A linear programming problem (LPP) is one
which optimizes (maximizes or minimizes) a linear function subject to a finite
collection of linear constraints. Formally, any LPP having  𝑛 decision variables
can be written in the following form:

n
Optimize Z = ∑ C j X j
j =1

Subject to
m

∑ a X (≤, =, ≥)b , i = 1, 2,..., m

i =1
ij j i

X j ≥ 0, j = 1, 2,..., n

where C j , aij , bi are constants.

Common terminology for the aforementioned linear programming model can now
be summarized as follows. The function, being optimized (maximized or
minimized), is referred to as the objective function. The restrictions normally are
referred to as constraints. The first m constraints (those with a function of all
the variables, on the left-hand side) are called functional constraints (or
structural constraints). Similarly, the X j ≥ 0, j = 1, 2,..., n restrictions are called
non-negativity constraints (or non-negativity conditions) and the aim is to find
the values of the variables Xj. Any vector 𝑋!  , satisfying the constraint of the
LPP is called a feasible solution of the problem (Fogiel, 1996; Schulze, 1998;
Chinneck, 2000). In this paper, we use the following terminology for the solution.
A feasible solution is a solution for which all the constraints are satisfied. An
infeasible solution is a solution for which at least one constraint is violated. The
problem is to find the values of the decision variables 𝑋! that maximize the
objective function 𝑍 subject to the m constraints and the non-negativity
restriction on the 𝑋! variable. The resulting set of decision variables that
maximize the objective function is called the optimal solution.

Results and Discussion

The information collected from the case company in addition to the sales and
other operating data was analyzed to provide estimates for LPP model
parameters. To set up the model, the first level decision variables on the volume
of products to be produced were set.
x1 = number of Polo T-shirts
x2 = number of basic T-shirts

Open  Science  Journal  –  June  2017     6  

 
Open  Science  Journal  
Research Article
 

x3 = number of mock neck T-shirts

x4 = number of singlets
x5 = number of short pants
Z = total profit during the month
Now, the linear programming model, maximizing the total profit is:

Maximize Z (Birr Profit) = 5.25 x1 + 4.82x2 + 4.85x3 + 3.2x4 + 6.75x5

Subject to
315x1 + 200x2 + 195x3 + 180x4 + 280x5 ≤ 38,665,000 (Fabric)
230x1 + 110x2 + 140x3 + 100x4 + 200x5 ≤ 26,638,120 (Thread)
12.5x1, +5.5 x2 + 6.3x3 + 4.25x4 + 7.5x5 ≤ 1,009,008 (Labor)
31x1 + 19.1x2 + 20.1x3 + 37.5x4 + 20.1x5 ≤ 4,979,414 (Over heads)
1.8x1 + 1.1x2 + 1.7x3 + 1.1x4 + 2.6x5 ≤ 346,200 (Cutting time)
22.7x1 + 5.4x2 + 10.4x3 + 4.5x4 + 40.1x5 ≤ 2,670,300 (Sewing time)
2x1 + 1.3x2 + 1.9x3 + 1.3x4 + 2.6x5 ≤ 371,700 (Finishing time)
x1≥ 15000; x2 ≥ 35800; x3 ≥ 13500; x4 ≥ 12500; x5 ≥ 16100 (Customer
orders)

Model Solution
A powerful linear programming problem solving technique is the simplex
method. Among the various software packages, LINGO 16.0 software was used to
hold the simplex procedures. The global optimal solution report for this model is
as follows.
Objective value: 777877.3
Infeasibilities: 0.000000
Total solver iterations: 2
Elapsed run time seconds: 0.03
Model class: LP
Total variables: 6
Nonlinear variables: 0
Integer variables: 0
Total constraints: 14
Nonlinear constraints: 0
Total nonzeros: 47
Nonlinear nonzeros: 0

Variable Value
Reduced Cost
Z 777877.3 0.000000
x1 15000.00 0.000000
x2 102287.8 0.000000
x3 13500.00 0.000000
x4 12500.00 0.000000
x5 16100.00 0.000000

Open  Science  Journal  –  June  2017     7  

 
Open  Science  Journal  
Research Article
 

Row Slack or Surplus Dual Price

1 777877.3 1.000000
2 0.000000 1.000000
3 4091936. 0.000000
4 5576460. 0.000000
5 0.000000 0.8763636
6 1897867. 0.000000
7 128123.4 0.000000
8 1246957. 0.000000
9 124965.8 0.000000
10 0.000000 5.704545
11 66487.82 0.000000
12 0.000000 0.6710909
13 0.000000 0.5245455
14 0.000000 0.3227273

Here, there was a difference between the LPP solutions obtained to satisfy
customer orders using LINGO 16.0 and actual production in Table 3. In the
former case, the product mix was Polo T-shirts, basic T-shirts, Mock neck T-
shirts, singlets, and short pants with volumes of 15,000.00, 102,287.8, 13,500.00,
12,500.00, and 16,100.00 respectively, and with a total profit of Birr 777,877.3 per
month upon selling. In the latter case, the product mix was Polo T-shirts, basic
T-shirts, mock neck T-shirts, singlets, and short pants with optimal volumes of
15,000.00, 35,800.00, 13,500.00, 12500.00, and 16,100.00 respectively, and with a
total profit of Birr 465,456 per month. At optimality, resources consumed by the
LINGO 16.0 software result were compared with the customer orders during the
month. In this case, the profit of the company could be improved by 59.84 %.
From Table 2, the monthly consumption values of customer orders for each
available resource were gathered from the company’s records. These consumption
values and LPP consumption values are summarized in Table 4. The ratios of
monthly consumption of the resources held were calculated to find the percentage
usage by each T-shirt style.

Table 4: Monthly consumption by LPP techniques and customer order

production.
Resources Held per Month Monthly Resources Percentage (%) of

Consumption Usage

Customer Customer
LPP LPP
Type Unit Value order order

Fabrics Gram 38665000 20,526,871 34573064 53.09

89.42

Threads Meter 26,638,120 13,288,700 21061660 49.89 79.07

Labor Birr 1,009,008 590,765 1,009,008 58.55 100

Open  Science  Journal  –  June  2017     8  

 
Open  Science  Journal  
Research Article
 

Overheads Birr 4979414 2,116,525 3081547 42.51 61.89

Cutting Min 346302 139,225 218076.2 40.20 62.97

Sewing Min 2670336 1,002,344 1423343 37.54 53.30

Finishing Min 371628 154,982 246734.2 41.70 66.38

The study shows that the LPP resource utilization of fabrics, threads, labor,
overheads, cutting, sewing, and finishing can be significantly improved to 89.42%,
79.07%, 100%, 61.89%, 62.97%, 53.30%, and 66.38% respectively. Thus, Figure 1
shows that production based on customer orders lead to inefficient resource
utilization since most of the resources are idle.

100!
90!
80!
70!
60!
50!
LPP!
40!
30!
20! Customer
10! Order!
0!

Figure 1: Comparison of customer order and LPP production resources

utilization.

Here, an analysis has been made without considering customer orders to

develop an LPP model using monthly consumption of resources. The monthly
consumption of each resource values are given under the left-hand side column in
Table 2, which can be used as required for the constraints.

Maximize z = 525/100*X1+482/100*X2+485/100*X3 +32/10*X4+675/100*X5;

Subject to
315*X1+ 200*X2 + 195*X3 + 180*X4+ 280*X5 ≤ 20526871;
230*X1 +110*X2+ 140*X3+ 100*X4+ 200*X5 ≤ 13288700;
125/10*X1+5.5 *X2 + 63/10*X3+ 425/100*X4 + 75/10*X5 ≤ 590765;
31/10*X1+ 191/10*X2+ 201/10 *X3+ 165/10*X4+ 375/10*X5 ≤ 2116525;
18/10*X1+ 11/10*X2+ 17/10 *X3+ 11/10*X4+ 26/10*X5 ≤ 139225;
227/10*X1 + 55/10*X2+ 104/10*X3+ 45/10*X4 + 201/10*X5 ≤ 1002344;
2*X1+ 13/10*X2 + 19/10*X3+ 13/10*X4+ 26/10 *X5 ≤ 154982;
X1≥ 0; X2 ≥ 0; X3 ≥ 0; X4 ≥ 0; X5 ≥ 0.

Open  Science  Journal  –  June  2017     9  

 
Open  Science  Journal  
Research Article
 

The optimal solution values of the decision variables were obtained using the
LINGO 16.0 s software. Table 5 shows the monthly product mix of the actual
system obtained from the factory and that suggested by the LPP model. These
values have been multiplied by their respective unit profits to obtain the profit
per month of each product. In this case where customer orders were not
considered, the product mix suggested by the LPP model was Polo, basic T-
shirt, Mock neck T-shirts, singlets, and short pants at optimal volumes of 0,
330,538.4, 140,265.7, 0, and 28,254.43 respectively with a total profit of Birr
499,058.5 per month. It can be shown then that with the LPP optimal solution,
the profit of the company can be improved by 7.22% (499,058.5 –
465,456/465,456) ×100. Adopting operational research techniques in the
production decision help the company to improve its objective.

Table 5: Comparison of customer order production and LPP values.

Profit Production

per Unit Volume/Month Profit (Birr) /Month

(Birr) Customer Customer

Product Type Orders LPP Orders LPP

Polo T-shirts 5.25 15000 0 78,750 0

Basic T-shirts 4.82 35800 68576.43 172,556 330,538.4

Mock Neck T-

shirts 4.85 13500 28920.76 65,475 140,265.7

Singlets 3.2 12500 0 40,000 0

Short Pants 6.75 16100 4185.841 108,675 28,254.43

Total Profit/Month (Birr) 465,456 499,058.5

Conclusion

Ability to use resources (resource utilization) was recorded as the major

constraint in the apparel manufacturing industry. The profits comparison
between the actual production and production using LPP models show sizeable
differences. From this point of view, it can be concluded that the apparel
company should use quantitative research methods of linear programming to
determine their optimal product mix. Thus, it will be possible to obtain the
following results:
Ø The profit of the company can be improved by 59.84% (from Birr
465,456 per month to Birr 777,877.3 per month).
Ø Denying customer orders and adopting the LPP solution provides only
three types of products but the profit of the company can be improved
by 7.22% (from Birr 465,456 per month to Birr 499,058.5 per month).

Open  Science  Journal  –  June  2017     10  

 
Open  Science  Journal  
Research Article
 

Ø Use of an operational research technique in the production time horizon

helps the company to improve its objective.

References
Arefayne, D., & Pal, A. (2014). Productivity Improvement through Lean Manufacturing Tools: A
Case Study on Ethiopian Garment Industry. International Journal of Engineering Research &
Technology (IJERT), 3 (9), 1037–1045.
ETIDI. (2014). Textile Industry Development in Ethiopia. An Overview of Facts and Opportunities,
Addis Ababa, Ethiopia.
Ezema, B.I and Amakoml, U. (2012). Optimizing Profit with the Linear Programming Model: A
Focus on Golden Plastic Industry Limited, Enugu, Nigeria. Interdisciplinary Journal of
Research in Business, 2 ( 2), 37–49.
Kumar, V. (2010). JIT Based Quality Management: Concepts and Implications in Indian Context.
International Journal of Engineering Science and Technology, 2 (1), 40–50.
Reeb, J. and Leavengood, S. (1998). Using the Simplex Method to Solve Linear Programming
Maximization Problems. Oregon State University, Extension Service
Samuel, S. (2012). Investigation on the Effect of Supply Chain Integration on Ethiopian Garment
Industry’s Performance. A Thesis Submitted to the Department of Management. Addis Ababa
University, Addis Ababa, Ethiopia.
Shaheen, S. and Ahmad, T. (2015). Linear Programming Based Optimum Resource Utilization for
Manufacturing of Electronic Toys. International Research Journal of Engineering and
Technology (IRJET), 2 (1), 261–264
Yahya W. B. (2004). Determination of Optimum Product Mix at Minimum Raw Material Cost, Using
Linear programming. Nigeria Journal of Pure and Applied Sciences, 19, 1712–1721.

Open  Science  Journal  –  June  2017     11  

 
78 IJCSNS International Journal of Computer Science and Network Security, VOL.10 No.4, April 2010

A study of Transportation Problem for an Essential Item of Southern Part

of North Eastern Region of India as an OR Model and Use of Object
Oriented Programming

Nabendu Sen1, Tanmoy Som2 , Banashri Sinha1

1
Department of Mathematics, Assam University, Silchar; 2Department of Applied Mathematics, BHU, Varanasi, India

Summary
In this paper we formulate an OR model from the collected 2.2 Table 2
data concerning with the transportation of an essential item, X* Y* Z* U*
rice, from different suppliers of Silchar to different 6 12 7 18
destinations in Mizoram. In this study an attempt has been A*
0 0 5 0
made to analyze the optimal solution with basic feasible 5 10 6 16
solutions obtained using different methods [10]. Also B*
8 0 0 5
programs have been developed using object oriented 6 11 6 17
programming, C++. C*
2 0 5 0
Key words: 6 11 8 17
O.R model, feasible solution, VAM, object oriented D*
5 5 0 5
programming.. 7 13 8 19
1. Introduction E*
0 5 5 5
Mizoram, a part of southern region of North East India, not
being well connected from the other parts of the nation as N.B. i) A*, B*, C*, D*, E* are the code names of the
well as the North East region, depends on the market of the suppliers. ii) The transportation cost of different suppliers
adjacent district Silchar of Assam for its essential goods to the same destination varies to some extent due to their
like rice, flour, salt etc. Different suppliers of Silchar own policies.
regularly supply rice to the different markets of Mizoram.
[6-9]. As such the related data has been collected from the
The next table 3 shows the quantity available
concerned suppliers for the purpose of the mathematical
formulation. with these suppliers for a particular year.
2. Tables, Figures and Equations Table 3
2.1 Tables and Figures
Quantity Available
Table 1 shows the distance of different destinations in Supplier
(Quintal)
Mizoram from Silchar district.
Table1 A* 8000
Place Distance (Km)
B* 9200
Kolashib 90
Serchip 300 C* 6250
Aizwal 180 D* 4900
Saiha 450
E* 6100
We use the following code for the destination Kolashib,
Serchip, Aizwal, Saiha.
X* for Kolashib, Y* for Serchip, Z* for Aizwal and U* for
Saiha. The next table 4 shows the total demand of the destination
Transportation cost per quintal of rice effective from 2007 X*, Y*, Z*, U* from these suppliers during the year
from different suppliers to the different destinations (as
mentioned above) is displayed in the table 2.
IJCSNS International Journal of Computer Science and Network Security, VOL.10 No.4, April 2010 79

Table 4 V*
X* Y* Z* U* Availability
8000
Destination Demand (Quintals) A*
60 120 75 180 0
X* 5000 9200
B*
Y* 2000 58 100 60 165 0
6250
Z* 10000 C*
62 110 65 170 0
U* 6000 4900
D*
65 115 80 175 0
3. Methods of Obtaining Initial Basic Feasible 6100
Solutions:
We apply here the following methods to get the initial E* 0
basic feasible solutions: 70 135 85 195
(a) Northwest Corner Method (see [10]). 5000 2000 10000 6000 11450
(b) Vogel Approximation Method (see[10]). Demand
(c) Least Cost Method(see[10])
(d) Row Minima Method(see[10])
(e) Column Minima Method(see[10]) Now apply the three different methods for initial basic
4. Formulation of Model: feasible solution.
In this problem we make a transportation schedule for rice, Initial BASIC FEASIBLE SOLUTION BY
as being the essential commodity (main food of the people) NORTHWEST CORNER METHOD:
for the state of Mizoram. Combining the data of the tables After applying this method which leads to the following
2, 3 and 4, we get the following transportation model to final table as:
determine an optimal schedule so as to minimize the
transportation cost for rice to different markets of Mizoram.

X* Y* Z* U* Availability

A* 8000
60 120 75 180
B* 9200
58 100 60 165

C* 6250
62 110 65 170

D* 4900
65 115 80 175

E* 6100
70 135 85 195
Demand 5000 2000 10000 6000
From NWCM Method, we find number of occupied cell is
Here Σ a i = 34450, Σ b j = 23000. Since Σ a i ≠ Σ b j, we (5+5-1=9) which is exactly same as m+n-1. Therefore we
introduce a dummy destination V* with requirement of get the initial feasible solution as
11450 units and zero (0) transportation cost, as shown in x11 =5000, x12 =2000, x13 =1000,
the next table in the form of balanced transportation x23 =9000, x24 =200, x34 =5800,
problem. x35 =450, x45 =4900, x55 =6100.
Total T.C is Rs. 21, 74,000/
80 IJCSNS International Journal of Computer Science and Network Security, VOL.10 No.4, April 2010

INITIAL BASIC FEASIBLE SOLUTION BY VOGEL From Least cost Method, we find number of occupied cell
APPROXIMATION METHOD: is (5+5-1=9) which is exactly same as m+n-1. Therefore
we get the initial feasible solution as
x11 =5000, x13 =750, x15 =2250,
After applying this method which leads to the x25 =9200, x33 =6250, x42 =1900,
following final table as: x43 =3000. X52 =100, x54 =6000
Total T.C is Rs 24,04,500/
INITIAL BASIC FEASIBLE SOLUTION BY ROW
MINIMA METHOD:
After applying this method which leads to the following
final table as:

From Vogel Approximation Method, we find number of

occupied cell is (5+5-1=9) which is exactly same as m+n-
1. Therefore we get the initial feasible solution as
x11 =5000, x13 =3000, x22 =2000,
x23 =1200, x24 =6000, x33 =5800,
From Row minima methods, we find number of occupied
x35 =450, x45 =4900, x55 =6100.
cell is (5+5-1=9) which is exactly same as m+n-1.
Total T.C is Rs 12,73,000/
Therefore we get the initial feasible solution as
x15 =5000, x21 =5000 x23 =750,
INITIAL BASIC FEASIBLE SOLUTION BY LEAST
x25 =3450, x33 =6250. X42 =1900,
COST METHOD:
x43 =3000 ,X52 =100, x54 =6000
After applying this method which leads to the following
Total T.C is Rs 23,83,250/
final table as:
INITIAL BASIC FEASIBLE SOLUTION BY
COLUMN MINIMA METHOD:
After applying this method which leads to the following
final table as:
IJCSNS International Journal of Computer Science and Network Security, VOL.10 No.4, April 2010 81

From Row minima methods, we find number of occupied Since d15 is most positive, therefore cell (1, 5) enters the
cell is (5+5-1=9) which is exactly same as m+n-1. basis. We allocate an unknown quantity θ to this cell and
Therefore we get the initial feasible solution as identify a closed loop involving basic cells around this
x13 =1550, x14 =1100 x15 =750, entering cell. Now θ = min {450, 3000} = 450, so we drop
x21 =5000, x23 =4200. X32 =2000, cell (3, 5).
x33 =4250 ,X44 =4900, x55 =6000 Solving this we get the final optimal table after one
Total T.C is Rs22,10,000/ iteration as given below:

5. OPTIMALITY: Final optimal table:

Taking initial basic feasible solution due to Vogel‟s
approximation method ,we now proceed for optimality
using MODI method . Here we determine a set of ui and vj
starting with u1=0 and using the relation cij = ui + vj for
occupied basic cells as shown below.
c11= u1 + v1  60 = 0 + v1  v1 = 60, c13 = u1 + v3  75 = 0
+ v3  v3 = 75,
c22 = u2 + v2  100 = u2 + v2  v2 = 75, c23 = u2 + v3  60 = u2
+ 75  u2 = -15,
c24 = u2 + v4  165 = -15 + v4  v4 =180, c33 = u3 + v3  65 =
u3 + 75  u3 = -10,
c35 = u3 + v5 0 = -10 + v5  v5 = 10, c45 = u4 + v5  0 = u4
+10  u4 = -10 and
c55 = u5 + v5 0 = u5 +10  u5 = -10
We now find net evaluations for unoccupied cells by using the
relation dij = zij - cij
d12 =z12 – c12 = u1 +v2 – 120 = -5,d14 =z14 – c14 = u1 +v4 – 180 = 0,
d15 =z15 – c15 = u1 +v5 – 0 =10, d21 =z21 – c21 = u2 +v1 – 58 = -13,
d25 =z25 – c25 = u2 +v5 – 0 = -5, d31 =z31 – c31 =u3 +v1 – 62 = -12,
d32 =z32 – c32 = u3 +v2 – 110 = -5, d34 =z34 – c34 = u3 +v4 – 170 = 0,
d41 =z41 – c41 = u4 +v1 – 65 = -15,d42 =z42 – c42 = u4 +v2 – 115 = -10,
d43 =z43 – c43 = u4 +v3 – 80 = -15,d44 =z44 – c44 = u4 +v4 – 175 = -5,
d51 =z51 – c51 = u5 +v1 – 70 = -20,d52 =z52 – c52 = u5 +v2 – 135 = -30,
d53 =z53 – c53 =u5 +v3 – 85 = -20,d54 =z54 – c54 =u5 +v4 – 195 = -25

Then the initial iteration is given by Here all dij ≤ 0, so an optimal solution has reached as given
Initial Iteration: below:
x11 = 5000, x13 = 2550, x15 = 450,
x22 = 2000, x23 =1200, x24 = 6000,
x33 = 6250, x45 = 4900, x55 = 6100.
Thus the optimal transportation cost is Z = Rs.
12,46,000.00/

6. Pseudo Code for different methods for initial basic

feasible solution:
6.1 North West Corner Method-

define row_max = 5;
define col_max=5;
//create supply_array and require_array
float supply_array[row_max];
float require_array[col_max];
// creating the cost matrix and unit matrix
float cost_matrix[row_max][col_max];
float unit_matrix[row_max][col_max];
// initialize cost_matrix
for i:0 to row_max-1
for j:0 to col_max-1
cin>>cost_matrix[i][j];
end loop
82 IJCSNS International Journal of Computer Science and Network Security, VOL.10 No.4, April 2010

end loop end loop

// initialize unit_matrix //displaying the minimal cost
for i:0 to row_max-1 cout<<”the minimal cost obtained is :”<<cost_minimal;
for j:0 to col_max-1 end
unit_matrix[i[j]=0;
end loop 6. 2 Vogel Approximation Method-
end loop #define TRUE 1
float cost_minimal = 0.0; #define FALSE 0
float *supply_ptr; #define INFINITY 1111
float *require_ptr; #define N 3
supply_ptr = &supply_array[0]; #define M 4
require_ptr = &require_array[0]; void input(void);
// initialize supply_array void display(void);
for i:0 to row_max-1 void displayfinal(void);
cin>>supply_array; void diffmin(void);
end loop void table(void);
//initialize require_array int max(int *,int *,int);
for i:0 to col_max-1; int min(int,int);
cin>>require_array; int mini(int *,int *,int);
float *matrix_ptr; int condition(void);
matrix_ptr=&cost_matrix[0][0]; int arr[N][M];
int r = 0,c=0,x=0,y=0; int arrcopy[N][M];
while(x<=row_max-1 && y=col_max-1) int value[N][M];
{ int u[N];
if(*require_ptr>*supply_ptr) int v[M];
{ int rowdiffmin[N];
unit_matrix[x][y]=supply_array[x]; int coldiffmin[M];
require_array[y]=require_array[y]-unit_matrix[x][y]; int decide[M+N];
supply_array[x]= supply_array[x]-unit_matrix[x][y]; int x[N],y[M]; /* x is u y is v */
cost_minimal = cost_minimal+unit_matrix[x][y]; //main point of execution starts from here
x=x+1; { int i,j;
supply_ptr=supply_ptr+1; table(); x[0]=0;
matrix_ptr=matrix_ptr+col_max;
continue;
} for(i=0;i<3;i++)
if(*require_ptr<*supply_ptr) { for(j=0;j<4;j++)
{ { if(value[i][j]!=0)
unit_matrix[x][y]=require_array[y]; cout<<”U[i+1]+V[j+1]= arr[i][j])”;
require_array[y]=require_array[y]-unit_matrix[x][y]; }
supply_array[x] = supply_array[x]-unit_matrix[x][y]; }
cost_minimal = getch();}
cost_minimal+unit_matrix[x][y]*cost_matrix[x][y]; void table(void)
y=y+1; { int rowdiffminmaxpos;
require_ptr=require_ptr+1; int coldiffminmaxpos;
matrix_ptr=matrix_ptr+1; int decidemaxpos;
continue; int temp;
} int temparr[M];
if ( *require_ptr==*supply_ptr) int i;
{ clrscr();
unit_matrix[x][y]=require_array[y]; input();
require_array[y]=require_array[y]-unit_matrix[x][y]; diffmin();
supply_array[x]=supply_array[x]-unit_matrix[x][y]; display();
cost_minimal=cost_minimal+unit_matrix[x][y]*cost_matrix[x][y while(condition())
]; { max(decide,&decidemaxpos,M+N);
y=y+1; if(decidemaxpos>=0 && decidemaxpos<N)
x=x+1; { rowdiffminmaxpos=decidemaxpos;
require_ptr=require_ptr+1; for(i=0;i<M;i++)
supply_ptr=supply_ptr+1; temparr[i]=arr[decidemaxpos][i];
matrix_ptr=matrix_ptr+col_max; mini(temparr,&coldiffminmaxpos,M);
continue; }
} else if(decidemaxpos>=N && decidemaxpos<M+N)
} { coldiffminmaxpos=decidemaxpos-N;
// displaying the unit matrix for(i=0;i<N;i++)
for i:0 to row_max-1 temparr[i]=arr[decidemaxpos][i];
for j:0 ti col_max-1 temparr[i]=INFINITY;
cout<<unit_matrix[i][j]; mini(temparr,&rowdiffminmaxpos,M);
end loop }
IJCSNS International Journal of Computer Science and Network Security, VOL.10 No.4, April 2010 83

temp=min(u[rowdiffminmaxpos],v[coldiffminmaxpos]); 7 9 1|25 3|4 4 7|30 5 1111 1111 7

value[rowdiffminmaxpos][coldiffminmaxpos]=temp; 1111 30 (1104)
if(temp==u[rowdiffminmaxpos]) X[1][2] = 18
{ for(i=0;i<M;i++) X[1][3] = 1
arr[rowdiffminmaxpos][i]=INFINITY; X[2][1] = 12
u[rowdiffminmaxpos]-=temp; X[2][4] = 25
v[coldiffminmaxpos]-=temp; X[3][1] = 4
} X[3][3]=30
else if(temp==v[coldiffminmaxpos]) /**************************************************/}
{ for(i=0;i<N;i++) void displayfinal(void)
arr[i][coldiffminmaxpos]=INFINITY; { int i,j;
u[rowdiffminmaxpos]-=temp; cout<<endl;
v[coldiffminmaxpos]-=temp; for(i=0;i<N;i++)
} for(j=0;j<M;j++)
diffmin(); arr[i][j]=arrcopy[i][j];
getch(); for(i=0;i<N;i++)
display(); } { for(j=0;j<M;j++)
getch(); if(value[i][j]==0)
displayfinal(); cout<< arr[i][j];
getch(); else
} cout<< arr[i][j]<<”|”<< value[i][j];
void input(void) cout<<endl;
{ int i,j; }
for(i=0;i<N;i++) cout<<endl;
for(j=0;j<M;j++) for(i=0;i<N;i++)
arr[i][j]=arrcopy[i][j]=-1; /* Demand supply matrix */ for(j=0;j<M;j++)
arr[0][0]=arrcopy[0][0]=5; if(value[i][j]!=0)
arr[0][1]=arrcopy[0][1]=3; cout<<X[i+1][j+1]= value[i][j];}
arr[0][2]=arrcopy[0][2]=6; int condition(void)
arr[0][3]=arrcopy[0][3]=2; { int i; int flag; int temp[M+N]; flag=1;
arr[1][0]=arrcopy[1][0]=4; for(i=0;i<N;i++)
arr[1][1]=arrcopy[1][1]=7; temp[i]=u[i];
arr[1][2]=arrcopy[1][2]=9; for(;i<M+N;i++)
arr[1][3]=arrcopy[1][3]=1; temp[i]=v[i];
arr[2][0]=arrcopy[2][0]=3; for(i=0;i<M+N;i++)
arr[2][1]=arrcopy[2][1]=4; {
arr[2][2]=arrcopy[2][2]=7; if(temp[i]!=0)
arr[2][3]=arrcopy[2][3]=5; flag=0;
/* Supply */ }
u[0]=19; u[1]=37; u[2]=34; if(flag==0)
/* Demand */ v[0]=16; v[1]=18; v[2]=31; v[3]=25; return(TRUE);
/**************************************************\ 5 else
3 6 2 19 (1) 4 7 9 1 37 return(FALSE);
(3) 3 4 7 5 34 (1) 16 18 31 }
25 (1) (1) (1) (1) -------------------------------------------- int min(int a,int b)
-------- 5 3 6 1111 19 (2) 4 7 9 { if(a>b)
1111 12 (3) 3 4 7 1111 34 (1) return(b);
16 18 31 0 (1) (1) (1) (0) ----------------------- else
----------------------------- 5 3 6 1111 19 (2) return(a);}
1111 1111 1111 1111 0 (0) 3 4 7 int mini(int *a,int *aminpos,int n)
1111 34 (1) 4 18 31 0 (2) (1) (1) { int i; int amin;
(0) ---------------------------------------------------- 5 1111 amin=a[0]; *aminpos=0;
6 1111 1 (1) 1111 1111 1111 1111 for(i=0;i<n;i++)
0 (0) 3 1111 7 1111 34 (4) 4 0 { if(a[i]<amin)
31 0 (2) (0) (1) (0) -------------------------------------- { amin=a[i]; *aminpos=i; }
------------- 1111 1111 6 1111 1 (1105) }
1111 1111 1111 1111 0 (0) 1111 1111 7 return(amin);}
1111 30 (1104) 0 0 31 0 (0) (0) (1) int max(int *a,int *amaxpos,int n)
(0) --------------------------------------------------- 1111 1111 { int i; int amax; amax=a[0]; *amaxpos=0;
1111 1111 0 (0) 1111 1111 1111 1111 for(i=0;i<n;i++)
0 (0) 1111 1111 7 1111 30 (1104) 0 { if(a[i]>amax)
0 30 0 (0) (0) (1104) (0) ----------------------------- { amax=a[i]; *amaxpos=i; }
---------------------- 1111 1111 1111 1111 0 (0) }
1111 1111 1111 1111 0 (0) 1111 1111 return(amax);}
1111 1111 0 (0) 0 0 0 0 (0) (0) void diffmin(void)
(0) (0) --------------------------------------------------- ----------- { int min1,min2; int arrmin1pos,arrmin2pos; int i,j;
---------------------------------------- 5 3|18 6|1 2 4|12 for(i=0;i<N;i++)
84 IJCSNS International Journal of Computer Science and Network Security, VOL.10 No.4, April 2010

{ min1=arr[i][0]; arrmin1pos=0; cin>>cost_matrix[i][j];

for(j=0;j<M;j++) end loop
{ if(arr[i][j]<min1) end loop
{ min1=arr[i][j]; arrmin1pos=j; } i=0;
} if(arrmin1pos==1) j=0;
{ min2=arr[i][0]; arrmin2pos=0; } /* create unit_matrix & initialize it to zero
else float unit_matrix[R][C;
{ min2=arr[i][1]; arrmin2pos=1; } for i=0 to R-1;
for(j=0;j<M;j++) float j:0 to C-1
{ if(arr[i][j]<min2 && j!=arrmin1pos) unit_matrix[i][j]=0;
{ min2=arr[i][j]; end loop;
arrmin2pos=j; } } end loop;
rowdiffmin[i]=min2-min1; /* create supply_array and demand_array
decide[i]=rowdiffmin[i]; float supply_array[R];
} float require_array[C];
for(i=0;i<M;i++) float *cost_matrix_ptr;
{ min1=arr[0][i]; arrmin1pos=0; for(j=0;j<N;j++) cost_matrix_ptr = & cost_matrix[0][0];
{ if(arr[j][i]<min1) while(count<R-C+4)
{ min1=arr[j][i]; arrmin1pos=j; {
} float minr_array[C]=0;
} float minc_array[R]=0;
if(arrmin1pos==1) find_min_cost_matrix(cost_matrix[i][j]);
{ min2=arr[0][i]; arrmin2pos=0; struct matrix min_loc=find-min_loc(cost_matrix[i][j]);
} int a = i;
else int b = min_loc;
{ min2=arr[1][i]; arrmin2pos=1; int c=j;
} int x = 0;
for(j=0;j<N;j++) j=a;
{ if(require_array[min_loc.c]>supply_array[min_loc.r])
if(arr[j][i]<min2 && j!=arrmin1pos) {
{ min2=arr[j][i]; arrmin2pos=j; unit_matrix[min_loc.c][min_loc.c]= supply_matrix[min_loc.r];
} require_array[min_loc.c]=require_array[min_loc.c]-
} unit_matrix[min_loc.r][min_loc.c];
coldiffmin[i]=min2-min1; supply_array[min_loc.c]=supply_array[min_loc.r]-
decide[i+N]=coldiffmin[i]; unit_matrix[min_loc.r][min_loc.c];
} delete_row_costmatrix(cost_matrix[min_loc.r][min_loc.c]);
} // the row is deleted
void display(void) if(min_loc.r==0)
{ int i,j; cout<<endl; for(i=0;i<N;i++) i=i+1;
{ //construct new cost matrix(cost_matrix[i][j])
for(j=0;j<M;j++) count++;
cout<< arr[i][j]; float cost = cost +
cout<< u[i]<<””<< rowdiffmin[i]; printf("\n"); unit_matrix[min_loc.r][min_loc.c]*cost_matrix[min_loc.r][min_l
} oc.c];
out<<endl; delete_row_cost_matrix(cost_matrix[min_loc.r][min_loc.c]);
for(i=0;i<M;i++) construct new cost_matrix(cost_matrix[i[j]);
cout<< v[i]; }
cout<<endl; if (require_array[min_loc.c]<=supply_array[min_loc.r])
for(i=0;i<M;i++) {
cout<< coldiffmin[i]; unit_matrix[min_loc.r][min_loc.c]-require_array[min_loc.c];
require_array[min_loc.c]=require_array[min_loc.c]-
} unit_matrix[min_loc.r][min_loc.c];
end; supply_array[min_loc.c]=supply_array[min_loc.r]-
unit_matrix[min_loc.r][min_loc.c];
6.3 Least cost method – count++;
struct matrix float cost = cost +
{ unit_matrix[min_loc.r][min_loc.c]*cost_matrix[min_loc.r][min_l
int r; oc.c];
int c; if(min_loc.c==0)
}; j=j+1;
define row_max R delete_cost_matrix(cost_matrix[min_loc.r][min_loc.c]);
define col_max C; construct_new_cost_matrix(cost_matrix[i][j]);
matrix find_minloc(float*); }
/* create initial matrix */ // display the final cost
float cost_matrix[R][C]; cout<<”the final cost is :”<<cost;
for i:0 to R-1 //display the final unit matrix
for j:0 to C-1 int l,m;
IJCSNS International Journal of Computer Science and Network Security, VOL.10 No.4, April 2010 85

for l:0 to R-1 {

for m:0 to C-1 find 2_min_array(cost_matrix[i][j]);
cout<<unit_matrix[l][m]; int loc2= find loc2min_array(cost_matrix[i][j]);
End loop; unit_matrix[a][loc2]=0;
End loop; i=i+1;
j=1;
6.4 Row Minima Method- cost=cost+unit_matrix[a][b]*cost_matrix[a][b];
define row_max R; x=x+1;
define col_max C; continue;
/* create initial matrix }
float cost_matrix[R][C]; }
for i:0 to R-1 /* display unit_matrix
for y:0 to C-1 int l,m;
cin>>cost_matrix[i[j] for l:0 to R-1
end loop for m:0 to C-1
end loop cout<<unit_matrix[l][m];
i=0; /* display the final min_cost
j=0; cout<<”the final minimal cost is” <<cost;
/* Create unit-matrix and initialize it to zero
float unit_matrix[R][C]; 6.5 Column minima method –
for i:0 to R-1 Pseudo-code –
for j:0 to C-1 define row_max R;
unit_martix[i][j]=0; define col_max C;
end loop /* create initial cost-matrix
end loop float cost_matrix[R][C];
/* create supply array and demand array for i:0 to R-1
float supply_matrix{R]; for j:0 to C-1
float require_matrix[C]; cin>>cost_matrix[i][j];
float*cost_matrix_ptr; end loop
cost_matrix_ptr = &cost_matrix[0][0]; end loop
int count=0; i=0,j=0;
while(count<R-C+4)] /* create unit matrix & initialize it to zero
{ float unit_matrix[R][C];
float minr_array[C]=0; for i:0 to R-1
float minc_array[R]=0; for j:0 to C-1
create_minr_array(cost_matrix[i][j]); unit_matrix[i][j]=0;
find_minr_array(cost_matrix[i][j]); end loop
int min_loc = find_min_loc(cost_matrix[i][j]); end loop
int a = i; /* create supply_array & demand_array
int b = min_loc; float supply_array[R];
int c = j; float demand_array[C];
int r = 0; float *cost_matrix_ptr;
if (require_array[min_loc].c>supply_array[min_loc].r) cost_matrix_ptr = &cost_matrix[0][0];
{ int count=0;
int x = min_loc; int x=0;
int y = 0; while (count<R-C+4)
unit_matrix[a[b] = supply_array[y]; {
require_array[min_loc]=require_array[min_loc]- float minr_array[C] = 0;
unit_matrix[a][b]; float minc_array[R]=0;
supply_array[y]=supply_array[y]-unit_matrix[a][b]; create_minc_array(cost_matrix[i][j]);
i=i+1; find_minr_array(cost_matrix[i][j]);
y=I; int min_loc = find min_loc(cost_matrix[i][j]);
count++; a= min_loc;
cost=cost+unit_matrix[a][b]*cost_matrix[a][b]; b=j;
continue; int c = j;
} int x=0;
If(require_array[min_loc]<supply_array[y]) if ( require_array[y]>supply_array[min_loc])
{ {
unit_matrix[i][b]=require_array[min_loc]; unit_matrix[a][b] = supply_array[min_loc];
require_array[min_loc]=require_array[min_loc]- require_array[y] = require_array[y] – unit_matrix[a][b];
unit_matrix[i][b]; supply_array[min_loc] = supply_array[min_loc] –
supply_array[y]=supply_array[y]-unit_matrix[a][b]; unit_matrix[a][b];
cost=cost+unit_matrix[a][b]*cost_matrix[a][b]; j=j+1;
y=y+1; count++;
continue; cost = cost +unit _matrix[a][b]*cost_matrix[a][b];
} continue;
if (supply_array[0]==require_array[min_loc]&==0) }
86 IJCSNS International Journal of Computer Science and Network Security, VOL.10 No.4, April 2010
if (require_array[y]<supply_array[min_loc])
{ Transportation
unit_matrix[a][b] = require_array[j];
require_array[j] = require_array[j] – unit_matrix[a][b];
supply_array[min_loc] = supply_array[min_loc] – Transportation Problem. International Journal of
unit_matrix[a][b];
cost = cost +unit _matrix[a][b]*cost_matrix[a][b];
j=j+1;
continue;
} Transportation Related Problem of Southern Assam and Its
If ( supply_array[min_loc]==require_array[0])
{ [7] Sen, N. and Som, T. (2008b) Mathematical Modeling of
find loc2min_array(cost_matrix[i][j]);
int loc2 = find loc2min_array(cost_matrix[i][j]);
unit_matrix[loc2][b]= 0;
j=j+1;
i=j;
x=x+1;
cost = cost +unit _matrix[a][b]*cost_matrix[a][b];
} [9] Sen, N. (2008d) Modeling of Transportation Related
/* display unit_matrix
int l,m;
for l:0 to R-1
for m:0 to C-1
cout<<unit_matrix;
/* display the final min_cost
cout<<”the final minimal cost is”<<cost;
7. Conclusion.
If this above optimal schedule is adopted by the suppliers
of rice to Mizoram it would not only involve minimization
of the transportation cost but it would also minimize the
consumption of fuel in transporting the goods by the
different carriers on the other hand. The optimal solution
obtained in this present investigation shows much more
closeness with initial basic feasible solution obtained by
Vogel approximation methods. The comparison of optimal
solution have been made with other methods of finding
initial solutions and observe that Vogel‟s method give the
better initial feasible solutions which are closer to optimal
solution. The objected oriented programs using c++ have
been developed and the compared with computed results
for initial basic feasible solutions. The comparison shows
that the computed results tally with the results obtained
c++ programming. Pseudo code for said programs is given
for better understanding.
8. Acknowledgments
Authors are thankful to the suppliers for providing
necessary data in this regards. Authors are also thankful to
Mr. Panthadeep Bhattacharjee, Assam University, silchar
for providing technical support during this work.
References
[1] Appa, G.M. (1973) The Transportation Problem and Its
Variants. Operational Research Quarterly, vol, 24, nr.1,
pp.79-99.
[2] Gass, S.I. (1990) On Solving the Transportation Problem.
Journal of Operational Research, Vol.41,nr.4, pp. 291-297.
[3] Intrator ,J(1989) A note on the Optimal Solution of a
Problems. Asia –Pacific Journal of
Operation Research ,Vol.6.nr.2,pp.207-208.
[4] Patel, S.M (1992) Resolution of Closed Loop in
Management and Systems,Vol. 81.nr.1,pp35-46.
[5] Rao, S.S (1987) Optimization Theory and Applications.
Wiley Eastern Limited.
[6] Sen, N. and Som,T. (2008a) Mathematical Modeling of
Optimal Solution. AUJSc, Vol. 31, nr.1, pp. 22-27.
Transportation Related Fare Minimization Problem of
South-Assam and an Approach to Its Optimal Solution.
IJTM, Vol. 32, nr. 3, pp. 201-208.
[8] Sen, N. and Som, T. (2008c) Mathematical Modeling of
Transportation Related Problem of Southern Assam with an
Approach to Its Optimal Solution. ASR, Vol. 22, nr. 1, pp.
59-67.
Problems of Northeastern Region of India and Their
Solutions: A Case Study of Southern Part, PhD Thesis
submitted to Assam University, Silchar, India.
[10] Taha, H.A. (2006)Operation Research :An Introduction
“,Prentice Hall ,7th Editions5,USA.
[11] Wilson, A.G; Coelho,J. D; Macgill,S.M and Williams,
H.C.W.L(1981) Optimization in Locational and Transport
Analysis. John Willey and Sons.
Dr.Nabendu Sen is working as an Assistant
Professor, Department of Mathematics,
Assam University, Silchar (India). He has
received Ph.D (Mathematics) in 2008 from
Assam University, Silchar. His area of
Research includes Optimization and
Mathematical Modeling. He has published
several papers journals of good repute. He
has reviewed many research papers in reputed Journals.
Ms. Banashri Sinha is currently pursuing
her research work under the supervision of
Dr. Nabendu Sen in Assam University,
silchar (India) on Optimization. She has
completed her M.Sc (Mathematics) from
Guwahati University, Assam (India) in 2008.
Dr.Tanmoy Som is serving as a Professor in
the Department of Applied Mathematics,
Institute of Technology, Banaras Hindu
University, Varanasi, India. He was awarded
Ph.D. in Mathematics by Banaras Hindu
University, Varanasi in 1986. He has
published more than 55 research papers in
the areas of „Functional Analysis,
Optimization Modeling, Fuzzy Set Theory and Image Processing
in the Journals of National and International reputes. He is also a
reviewer of many National and International Journals and
member of many academic bodies. He has delivered several
invited talks at National as well as International level Seminars &
Conferences. He has produced a good number of M. Phil. and
Ph.D. s.