ISSN (Print) : 0974-6846

Indian Journal of Science and Technology, Vol 9(39), DOI: 10.17485/ijst/2016/v9i39/100788, October 2016 ISSN (Online) : 0974-5645

Sensitivity Analysis and Optimal Production

Scheduling as a Dual Phase Simplex Model
Saurav Dhand and Ashish Singla*
Department of Mechanical Engineering, Thapar University, Patiala- 147001, Punjab, India;
[email protected]

Abstract
Background/Objectives: Production efficiency is mainly influenced by two factors ─ satisfying customer demands and
profit maximization. Some methods for optimizing the production are proposed in the literature, which are ­computationally
expensive. Methods/Statistical Analysis: Production scheduling problem is addressed by using a single objective ­function
and suitable operational constraints. Dual-Phase Simplex method is used to determine the optimal schedule for ­production
and sales. As compared to conventional methods, the current method is found to be computationally inexpensive and easy
to implement. The sensitivity analysis is performed to study the effects of parametric variations on the production ­volume.
Findings: The performance of a quality management system, in terms of production efficiency of an ­organization, is based
on the quality of decisions taken at the shop floor. A good production manager must be able to foresee all decision-related
discrepancies and adopt preventive measures appropriately. In order to resolve this contradictory requirement, an
­optimized production scheduling process is generally required, which deals with desired type and quantity of output
at minimum cost. A typical case study of Tema (Ghana) based cable manufacturing firm is considered to optimize the
­production scheduling problem. In this paper, a mathematical model for the optimizing the production of different items
having different costs is developed by considering the problem as Linear Programming Problem (LPP). Further, the formed
LPP contains mixed type constraints, which are solved using the Dual-Phase Simplex method. It has been found from the
literature that the Dual-Phase Simplex method is computationally inexpensive. Moreover, sensitivity analysis is done to
observe the effect of break-downs on the production rate. Application/Improvements: The idea proposed in this paper
has a wide scope in various physical processes to optimize their parameter values. However, for multi-objective problems
the Dual-Phase Simplex method gives invalid results.

Keywords: Production Scheduling, Simplex Method

1. Introduction seeks to discuss the application of the optimization concept

at production scheduling with the help of single objective
Production scheduling is considered as one of the ­oldest and typical operational constraints. In order to address
and hardest problems of the manufacturing systems. Since the production scheduling process of various industries, a
the scheduling problem is dynamic in nature, finding the number of powerful models are reported in the literature,
optimal solution is a tough task. However, with certain which are mainly based on linear programming models.
assumptions and given conditions, solution of the sched- For example, in yogurt producing industry, there is a wide
uling problem does not vary much from one system of variety of products according to different features like fat
operations to another. The main motive behind the pro- content, usage of whey to produce the mixture, the fla-
duction scheduling optimization is to reduce production vor, the size of the container or the label language. In this
costs, time and maximize the profit. An optimized pro- type of operations related to scheduling, constraints like
duction scheduling process is characterized by achieving satisfaction of multiple clients, variable processing times,
the desired output of products, both in type and quantity, setting up times according to sequence, production costs
within the planned time at minimum costs1.This paper hikes and efforts to keep an eye inventory levels need to

*Author for correspondence

 Sensitivity Analysis and Optimal Production Scheduling as a Dual Phase Simplex Model

be deliberately handled in order to retain the required 2012. This study covers the use of linear programming to
production objectives2. Optimal scheduling refers to a find the optimal solution for scheduling of production
schedule, which yields the right quantity and type within and sales. The study in this paper treats typical schedul-
the minimum expected time with the available resources ing problems in manufacturing companies with special
to maximize the profit. reference to a case study at Nexans Kabel metal Ghana
Optimal schedule consists of timely provision of input Limited, a production firm engaged in the production of
raw materials, identification of the appropriate machines house wiring cables12. The constraints to the production
for the job and provision of sufficient work instructions and sales capacities were based on the company’s activities
to operators. In this paper, the problem of a Ghana based over the period January-December, 2012. The company
cable manufacturing firm is considered for optimal sched- manufactures more than one hundred and fifty types of
uling. In order to meet market demands and the targeted Copper products on production program. However, an
sales as per the marketing strategy, an optimized produc- optimal scheduling solution need to be identified such
tion scheduling output model needs to be developed. Most that, at least 70 % of the total Copper sales can proceed
of industrial problems, like production scheduling can be from the six fast selling products.
expressed in mathematical form in order to optimize it. It In this paper, the Dual-phase Simplex method is used
has been observed in the past that predominant production to optimize the production scheduling process, which
scheduling problems falls in one of the ­following types: is not reported earlier. As compared to the previous
method like Simplex method, Big M method12, the pro-
• To avoid long term delivery schedules, companies
posed method is computationally inexpensive and easy
­prefer to satisfy the demands instantly.
to implement. Furthermore, sensitivity analysis is also
• After mapping of customer performance to product
done to check the effect of parameter variations on the
specification, shortage of one product would affect the
­production output.
sales pattern.
• Reference benchmark is needed to measure and
­analyze the performance at end of each month. 2. Structure of the Optimization
Production scheduling problem has gained the attention Problem
of many researchers in recent years. A fuzzy weighted
averages approach3has been used to describe non-linear Production scheduling problems can be expressed with
problems into simple extended algebric operations on one basic objective function subjected to a number of
fuzzy members. Fuzzy sets are represented using α-sets. constraints. Therefore, these kinds of problems can be
This method generates discrete but exact solution using ‘ treated as linear optimization problem with all the vari-
’ evaluations per α-cut. An efficient method4 improved ables taking real values. The structure of a general LPP
this existing approach using steepest-descent method to problem consists of a linear objective function of ‘N’
give solution in ‘ ’ iterations for each α-cut. The real variables is given as:
computational requirements are further optimized by  (1)
transforming the given non-linear problem into linear
where are ‘N’ real numbers. The
one5using Charnes and Cooper transformation method6.A
­constraints can be represented as:
detailed review7 discusses approximately 90 research
papers on multi-objective scheduling process involving  (2)
Genetic Algorithm8, Ant-Colony optimization9, Particle
Swarm algorithm10, Heuristics approach11etc. Different  (3)
approaches are categorized according to the application in
and  (4)
scheduling e.g. considering processing time, breakdown,
limited stock, job-shop scheduling, flow-shop scheduling where, are inequality constraints and
problem etc. In this paper, a case study of a cable manu- are equality constraints on objective function . If
facturing company is discussed, so as to achieve a monthly satisfy all the constraints of formulated
target of 92 tons of Copper to be manufactured and sold. LPP, then these are treated as feasible solutions. Further,
The target resulted from 14.6% increase in production in objective function and constraints are formulated in terms

2 Vol 9 (39) | October 2016 | www.indjst.org Indian Journal of Science and Technology
 Saurav Dhand and Ashish Singla

of the variables and finally this optimization ­problem • Constraint 1

can be solved using any linear programming method. It has been observed by the firm that, doubling the pro-
Commercial solvers are available to solve optimiza- duction of NCY1.5 RS and NCY2.5 RS, provided all
tion algorithms. However, a manual method, known as other production rates same, produces 10 tons extra than
Dual-Phase Simplex method, is considered in this work. expected target of 65 tons/month. i.e.
Dual-phase Simplex method is generally the ­extension
of Simplex method, which cannot be used for mixed con-
straint problems. For these kinds of problems, artificial
• Constraint 2
variables are used along with the slack variables.
Data obtained from the firm illustrates that, the copper
This method consists of following steps:
sales for 2012 was over 1000 tons, when it was targeted
• Adding slack variables near 960 tons. Following the same trend, the expected
• Addition of artificial variables sale till December end was decided to increase upto 1100
• Formulation of a dummy objective function to tonnes. Monthly expected sales was 92 tons. It was also
­calculate the initial basic feasible solution (Phase-1) observed that in 2012, the house-wiring cables contribute
• Using initial basic feasible solution to solve original 70 % of the total sales. Therefore, expecting the same con-
objective function using Simplex method tribution for 2013 also, monthly sales due to house-wiring
• Obtaining the optimal solution cables is 65 tons/month. i.e.

3. Problem Formulation
• Constraint 3
Dual-Phase Simplex method is considered in ­analyzing
From previous year’s sales pattern, management deduced
the data obtained12 from a firm, Nexans Kabel metal
that, by targeting more hotels and institutional building
Ghana Limited, manufacturer of electrical cable wires and
projects in order to sale more NCY4 RS and NCY6 RS,
conductors. With a 14.6% increased target of 2012-year
performance, in terms of production output and sales additional 23 tons are expected to be sold above previous
tonnage, an estimated target performance in 2013 is pro- 65 tons. This implies:
jected to 1100 tons of copper in 2013. In this case study,
the objective is to optimize the consumption of copper
used for producing different pairs of cables. In short, opti- Therefore, scheduling problem can be formulated as:
mizing Copper consumption gives an idea of quantity of
different cables to be produced. If: Maximize (5)
Optimal Copper consumption to produce Subject to:
=
different sets of wires;
Total output for NCY1.5 RS and NCY2.5 RS in
=
tons/month;
Total output for NCY4 RS and NCY6 RS in
=
tons/month;
Total output for NCY10 RS and NCY16 RS in Now following the Dual-Phase Simplex Method,
=
tons/month, as shown in (Table 1.)
which states that:
Table 1. Details of Production
3.1 Introduction of Slack/Surplus and
S. No. Type Purpose Quantity
NCY1.5 RS and Lighting and
Artificial Variables
1
NCY2.5 RS socket wiring First of all, the slack variables are added to all the lagging
NCY4 RS and Refrigeration and constraints. Similarly, the surplus variables and artificial
2
NCY6 RS A.C. variables are added to all the leading constraints. In addi-
NCY10 RS and tion to this, the artificial variables are added to equality
3 High voltage wires
NCY16 RS constraints also.

Vol 9 (39) | October 2016 | www.indjst.org Indian Journal of Science and Technology 3
 Sensitivity Analysis and Optimal Production Scheduling as a Dual Phase Simplex Model

Maximize (6) function value. Going on with the same procedure, the
function value is maximized to zero in Table 3. Since all
Subject to:
the optimality values are zero, hence this is the final solu-
tion for Phase 1. According to the Dual-Phase Simplex
method the solution of first phase can be treated as the
initial basic feasible solution for the Phase 2.

3.3 Phase 2: Finding Optima of Actual

3.2 Phase 1: Formulation of Dummy
Objective Function
Objective Function
From the Table 3 of Phase 1, it is clear that no artificial
Due to the presence of both equality and inequality
variable is now present in the system as basic variable.
­constraints, the dummy objective function (W) must be
Therefore the obtained initial basic feasible solution is
created for the initial basic feasible solution.
further analyzed using original objective function as
Maximize (7)
shown in Table 4. The function value at the starting of the
Phase 2 is 55. Performing the calculations using Simplex
Subject to:
algorithm, optimized function value comes out to be 65.
Simplex method is again applied to find the optimal value
for Phase 2. At optima:

Since, artificial variables cannot be treated as ­non-basic

variables, choose: From the above calculations, it can be concluded that
total optimum output for NCY1.5 RS and NCY2.5 RS is 10
tons/month. Similarly, total optimum output for NCY4 RS
which leads to basic variables as:
Table 3. Final Step for Phase 1
0 0 0 0 0
This is the initial basic feasible solution for the Phase
b
1 of the problem. It can be seen from the Table 2 that the
function value for the initial basic feasible solution is –153. 0 0.8 0 0 1 0 2.5
This much negative value of objective function is due to 0 1 0 1 0 1 42
the presence of negative artificial variables. This func-
0 0 1 0 0 –1 23
tion value is further optimized, using Simplex algorithm,
where it is found to be -21. It can be clearly seen that, only 0 0 0 0 0 0
after first iteration there is significant maximization in the 0 0 0 0 0

Table 2. First Step for Phase 1

Table 4. First Step for Phase 2
0 0 0 0 0 –1 –1
0 0 0 0 0
b θ
b θ
0 2 1 1 1 0 0 0 75 75 –1 0.8 0 0 1 0 2.5 3.125
–1 1 1 1 0 0 1 0 65 65 1 1 0 1 0 1 42 42
–1 1 2 1 0 –1 0 1 88 44 1 0 1 0 0 –1 23 –
–2 –3 –2 0 1 –1 –1 –153 0 1 1 –1 0 62.5

2 3 2 0 –1 0 0 1 0 0 0 –1 0

4 Vol 9 (39) | October 2016 | www.indjst.org Indian Journal of Science and Technology
 Saurav Dhand and Ashish Singla

and NCY6 RS is 23 tons/month and total output for NCY10 remain constant throughout the year. The projected
RS and NCY16 RS is 32 tons/month. Corresponding to ­performance obtained from solving the mathematical
all these optimum outputs, the maximum value of total model can however, be affected by changes depending
plant output comes out to 65 tons/months. Moreover, the upon a number of constraint factors. These operational
obtained optima also satisfy the given constraints. It was limitations may be dynamic in nature, and can have direct
observed from the sales of 2012 that sales and production bearing on scheduling outcomes of the management
of NCY1.5 RS is 70 % of that of NCY2.5 RS, NCY4 RS is ­system. These include:
54% of NCY6 RS and NCY10 RS is 70 % of total sales and
• Effect of time on production costs e.g. water, ­electricity
production of NCY16 RS12.
and labour changes in tariffs overtime
Let, total output from NCY1.5 RS = a; and Total ­output
• Stochastic nature of the expected sales volume over
from NCY2.5 RS = b; given,
the time, due to changes in price and other economic
developments
• Fluctuating costs of raw materials on the world market
and likelihood of shipment delays coupled with man-
agement’s obligations to maintain constant supply of
appropriate raw materials for production
• Preventive maintenance and machine breakdown
Similarly, if total output from NCY4 RS= c; and Total hours on production capacity
output from NCY6 RS = d;given, These factors may considerably influence the optimality
of the linear model. To examine the extent of deviations
of the actual scheduling output from the projected opti-
mal performance with regard to the impact of the ignored
Given limitations mentioned above, original data needs to be
altered, in systematic order. This process is well known
as sensitivity analysis. Sensitivity analysis is necessary for
following reasons:
• To test the reliability of the results of the model in case
Let, total output from NCY10 RS= e; and Total output of significant changes due to constraint factors
from NCY16 RS = f;given, • To show the relationship that exist between input and
output variables in the model
• To make recommendations more credible and
­persuasive to decision makers
Given Let the expected production schedules for NCY1.5
RS and NCY2.5 RS are affected by sudden machine

Table 5. Optimal solution of Original Problem (O.P.)

and Sensitivity Analysis (S.A.)
S. No. Product Qty (O.P.) Qty (S.A.)
The individual optimum output of all kinds of
­products is listed in Table 5. It is clear that this much out- 1 NCY1.5 RS 4.12 1.287
put of each product, results in the optimal total output 2 NCY2.5 RS 5.88 1.838
from the ­factory. 3 NCY4 RS 8.06 8.06
4 NCY6 RS 14.94 14.94
4. Sensitivity Analysis 5 NCY10 RS 13.18 16.008
6 NCY16 RS 18.82 22.867
It has been assumed that the design parameters that are
used to develop the model of scheduling problem will Total 65 65

Vol 9 (39) | October 2016 | www.indjst.org Indian Journal of Science and Technology 5
 Sensitivity Analysis and Optimal Production Scheduling as a Dual Phase Simplex Model

­ reakdown for several days beyond the promised repair

b requires the production data to be quantified and to
duration. This circumstance can delay customer delivery ­present in the form of linear relations. Although there are
schedules and reduce subsequent sales for the month. some limitations related to linear programming method
Assuming sales is reduced by 10%(depending on break- like restrictive use and non-applicability to non-linear
down hours), this will have effect on the constraints. problems, it gives very close result to the actual optimal
Demand for the two products shall reduce to 90%. We solution. In the last section of the paper i.e. sensitivity
can therefore alter the coefficient of and reduce also analysis, the effect of machine breakdown on output of
the expected sales tonnage by 10% to obtain 67.5 tons each product has been observed.
as the right hand side constant for same constraint set.
All other constraints remain the same as in the original
problem. 5. References
Maximize (5) 1. Doganis, P, Sarimveis H. Optimal production ­scheduling
for the dairy industry. Annals of Operations Research.
Subject to: 2008; 159(1):315–31.
2. Graves SC. A review of production scheduling. Operations
Research. 1981; 29(4):646–75.
3. Dong WM, Wong FS. Fuzzy weighted averages and
­implementation of the extension principle. Fuzzy Sets and
Systems. 1987; 21(2):183–99.
Proceeding in same manner as that of first case,
4. Liou TS, Wang MJ. Fuzzy weighted average: An improved
­adding slack, surplus, artificial variables and solving in algorithm. Fuzzy Sets and Systems. 1992; 49(3):307–15.
two phases, solution comes out to be: 5. Guh YY, Hon CC, Lee ES. Fuzzy weighted average: The
­linear programming approach via Charnes and Cooper’s
rule. Fuzzy Sets and Systems. 2001; 117(1):157–60.
The optimum results clearly depicts that the ­breakdown 6. Charnes A, Cooper WW. Programming with linear
of the machine breakdown has almost zero effect on the ­fractional functionals. Naval Research logistics quarterly.
output of i.e. total output for NCY4 RS and NCY6 RS 1962; 9(3–4):181–6.
7. Lei D. Multi-objective production scheduling: A survey.
as shown in Table 5, whereas, the total output for NCY10
The International Journal of Advanced Manufacturing
RS and NCY16 RS is increased to 38.875 with decrease in
Technology. 2009; 43(9–10):926–38.
total output for NCY1.5 RS and NCY2.5 RS. Hence, the
8. Jolai F, Rabbani M, Amalnick S, Dabaghi A, Dehghan
machine breakdown has no effect on the total output. M, Parast M Y. Genetic algorithm for bi-criteria single
machine scheduling problem of minimizing maximum ear-
5. Conclusion liness and number of tardy jobs. Applied Mathematics and
Computation. 2007; 194(2):552–60.
This paper has presented a mathematical model for the 9. T’kindt V, Monmarche N, Tercinet F, Laugt D. An ant
optimization of the production scheduling of different colony optimization algorithm to solve a 2-machine bicri-
items having different costs. The problem statements pre- teria flowshop scheduling problem. European Journal of
sented in this paper have been appropriately addressed, Operational Research. 2002; 142(2):250–7.
10. Lie D. A pareto archive particle swarm optimization for
with reference to the case study recommendations and
multi-objective job shop scheduling. Computers and
sensitivity analysis. The main objective of this study is to
Industrial Engineering. 2008; 54(4):960–71.
optimize the overall production of three sets of products
11. Chen W. An Efficient Algorithm for Scheduling Jobs on
using some constraints obtained from the factory data a Machine with Periodic Maintenance. The International
and availability of machines. The data used for the case Journal of Advanced Manufacturing Technology. 2007;
study was the average values obtained from the previ- 34(11–12); 1173–82.
ous year’s production. A linear programming approach 12. Addy I. optimization of production scheduling as linear
has been used for optimizing the output. This approach programming model. Ph.D. Thesis; 2013.

6 Vol 9 (39) | October 2016 | www.indjst.org Indian Journal of Science and Technology