Proceedings of 2017 4th International Conference on
Control, Decision and Information Technologies (CoDIT'17) / April 5-7, 2017, Barcelona, Spain
Applying Genetic Algorithm for Hybrid Job Shop
Scheduling in a Cosmetic Industry
Edgar Macias Fabricio Niebles
Int. School of Economic and Institute of Information
Management Sciences Systems (IWI),
Universidad de la Sabana, Universität Hamburg,
Bogotá, Colombia Hamburg, Germany
edgarmaal@unisabana.edu.co fabricio.niebles@uni-hamburg.de
Abstract— This work considers the problem of scheduling a
given set of jobs in a Flexible Job Shop in a cosmetic industry,
located in Colombia, taking into account the natural complexity
of the process and a lot of amount of variables involved, this
problem is considered as NP-hard in the strong sense. Therefore,
it is possible to find and optimal solution in a reasonable
computational time for only small instances, which in general,
does not reflect the industrial reality. For that reason, it is
proposed the use of metaheuristics as an alternative approach in
order to determine, with a low computational effort, the best
assignment of jobs in order to minimize the number of tardy
jobs. This optimization objective will allow to company to
improve their customer service. A Genetic Algorithm (GA) is
proposed. Computational experiments are carried out comparing
the proposed approach versus instances of literature by Chiang
and Fu. Results show the efficiency of our GA Algorithm.
Keywords—Scheduling; Genetic Algorithms; Cosmetic
Industry; Hybrid Jobshop.
I. Introduction
The new complex and changing dynamics of the current
productive environment, poses new challenges for companies,
who seek to increase simultaneously profits and customer
service level, by optimizing their resources with the aim of
being sustainable in the market, which implies greater
flexibility in production systems, from the inputs of the
productive chain, workflows, methods of planning and control
of production, among others aspects, that affect the quality of
the products, the satisfaction of the changing demand and the
personalized products (mass customization of production) [1];
In conclusion, in different productive sectors offering a variety
of products to fit the requirements and needs of customers, is
the goal of organizations [2].
This customization of production has allowed the
development of flexible production systems, which makes
complex one of the most important and critical processes
within production management, such as the scheduling
process [3], Which consists in sequencing of tasks (jobs)
taking into account the availability of the production resources
and the constraints of the productive system with the aim, to
optimize one or several objectives such as: Makespan, the
978-1-5090-6465-6/17/$31.00 ©2017 IEEE
Genett Jimenez Dionicio Neira
Dept. of Industrial Process Dept. of Industrial
Engineering, Engineering,
Institucion Universitaria ITSA, Universidad de la Costa CUC,
Soledad - Atlántico, Colombia Barranquilla, Colombia
gjimenez@itsa.edu.co dneira1@cuc.edu.co
number of tardy jobs, the Total tardiness, among others [4],
[5]. Other possible objective functions found in the literature
are maximum tardiness or total flow time [6].
Scheduling process is complex in a natural way due to the
variability in demand and the high variety of products, which
makes production planning based on historical data is an
inefficient process, due to the inconsistency of data between
one product and another, with decreases the productivity
and/or uses of available resources based on the configuration
of the productive system. One of the more complex production
systems used in industries such as cosmetics, object of this
work, is the type of Hybrid or Flexible Job Shop (FJSSP), in
which small lots are produced, which allows manufacturing a
large variety of customized products, with different volumes
of production, according to the customer requirements. This
type of configuration works with flexible resources, which are
grouped in workstations based on the working process [7],
which implies the use of procedures for scheduling to improve
the level of customer service, comply with the delivery dates
of orders, the minimization of delayed jobs, waiting times and
idle work times, which affect sales and industrial
competitiveness. For the above, the literature for the FJSSP,
which is a variation of the original job shop problem, is less
studied in the literature, but of great interest in the research
and application in manufacturing environments such as
apparel industry, furniture production, and cosmetic industry,
among others [8]. The FJSSP consists of planning and
organizing a set of tasks, which must be processed using a set
of resources with operative restrictions [7]. Some authors have
categorized it as a more complex problem than the traditional
job shop, taking into account the need for the allocation of the
operations [9].
Taking into account the natural complexity of the flexible
job shop configuration, this problem has been shown to be
NP-hard [10], [11]. Therefore, it is possible to find an optimal
solution using exact methods in a reasonable calculation time
only for small-sized instances which, in general, do not reflect
the industrial reality. For this reason, it is found in the
literature the use of heuristic and metaheuristic methods, as
alternative approaches for this problem, in order to determine,
with a low computational effort, solutions that are as close as
possible to the optimum value [6].
1015 -1-
 Proceedings of 2017 4th International Conference on
Control, Decision and Information Technologies (CoDIT'17) / April 5-7, 2017, Barcelona, Spain
Therefore, the present proposal aims to analyze the
problem of scheduling a certain set of jobs in a flexible job
shop configuration, taking as a case study the cosmetics and
toiletries sector, taking into account that it has the
characteristics of the job shop system, as well as for the
growing of this sector and international projection in the
Colombian economy [12], proposing the use of metaheuristic
approach, in order to minimize the number of tardy jobs and,
therefore, non-compliance for customers. This objective will
allow the improvement of the customer service level, as well
as the increasing of sales of the products.
This paper is structured as follows: In section 2 a brief
introduction and the literature review for the Flexible Job
Shop scheduling problem. Afterward, the context of the
cosmetic sector in Colombia and the problem description is
presented in Section 3. Section 4 describes the problem
formulation, with the mathematical model and the proposed
genetic algorithm for Flexible Job Shop configuration.
Validation and experimentation of Genetic Algorithm and
analysis of results are discussed in Section 5. Finally, Section
6 concludes the paper with some directions for future research.
Hybrid Job Shop Scheduling
II.
Problem: A Literature Review
The scientific literature has reported some academic works
and real-life applications regarding the utilization of
metaheuristics for solving the Hybrid or Flexible Job Shop
Scheduling problem, although less studied than the traditional
job-shop scheduling problem. Dedieu [14], addresses the
problem general of multiple lot sizing (job-shop) for
adaptation in manufacturing plants. Cruz et al. [15], worked
on an algorithm of satisfiability for the Job Shop scheduling
problem; Peña and Lillo [16], made a survey of Job Shop
Scheduling Problem; Mailing [17], used heuristic algorithms
and exact methods for the problem of job shop scheduling;
Seville and Zurita [18], proposed the use of metaheuristics for
the optimization of the sequence of production and the
allocation of labor in a manufacturing company; Moslehi and
Mahnam [19] worked on a Pareto approach for multi-objective
Job-shop scheduling problem with particle swarm
optimization and local search; Sun et al. [20], worked on a
Solving Job Shop Scheduling Problem Using Genetic
Algorithm with Penalty Function; Sels et al [21] developed a
population hybrid of one and two search procedures for Job-
shop scheduling problem; Drobouchevitch and Strusevich
[22], worked on a two-stage Heuristic of the Job-shop
scheduling problem with a bottleneck machine; Xiaohong et al
[23], worked on a research on Job-shop scheduling problem
based on the genetic algorithm. To the best of our knowledge
is the first time that Genetic Algorithm is applied to solving
the scheduling problem in a Cosmetic Industry with the
configuration of type Hybrid Job Shop. For that reason, this
paper represents a relevant contribution to operations research
and operations management field.
III. Problem Description
Cosmetics and toiletries, industry constitutes one of the
links in the Colombian economy, sustained growth in recent
years [24, 25, 26], in the “Fig. 1”, sales statistics are observed,
from 2009 to 2015, and projected to 2019 where is observed
the high potential of the sector.
Fig. 1. Colombian market cosmetics and toiletries 2000 - 2019P (Sales in
Million US $) ** P: Projected. Source © Euromonitor International 2016
To achieve this sustained growth, companies in the sector
must respond quickly changing requirements and market
demand, the "mass customization" seeking innovation and
added value based on the client [12] [24], [25]. This sector has
the characteristics high variety of products with different
volumes of production between each one of them, leading to
implementing systems such flexible manufacturing job shop,
with which to respond to the high variety of products and
changes in demand, leading to improving procedures for
production scheduling.
For the company studied, the following problems were
detected:
• High variability in demand and sales, by product
type, this makes the production planning based on
historical data, timely response to changes in demand,
affecting productivity and appropriate use of
resources
• The production system used by the cosmetic industry
analyzed is the job-shop type, where the material
flow is irregular, random and varies considerably
from one order to the next.
• The cosmetics industry chosen for the study consists
of 6 major product lines where it has been based
studio wooden pencils; for its large annual share sales
and the large share of machines that requires
processing this product, which becomes more
complex production control, optimization of
resources and compliance management indicators.
1016 -2-
 Proceedings of 2017 4th International Conference on
Control, Decision and Information Technologies (CoDIT'17) / April 5-7, 2017, Barcelona, Spain

• The company does not have software tools for Therefore,

production scheduling or formal procedures that meet
I {Pencil 1, 2,….., 10};
the needs of the Job-Shop process, variability in
demand or the installed capacity of the production J {Machines SC-1, SC-2,…, ET-2}; 18 Machines
process, generating a large number of tardy jobs, and
failure in deliveries to customers. The model variables are shown below:

In Table I, are observed the lines of cosmetic products pij = (Procedure time job “i” on machine “j)
manufactured by the company in the sector studied for this stij = (Start time of job i on machine j)
research. It was taken the wooden pencils cosmetics line, by
sales volume accounting for the company, and by the use that di = (date of delivery of job i)
most of the resources of the company. In addition, it presented Sij kl = (Setup time when it ends a process the pencil “I” in
a percentage higher than 30% of production that is not the machine “j” for a new operation is made)
delivered on time, increasing the cost for the company, the
number of tardy jobs and declining sales. Ci = (Job completion time “i”)
Ui = (Tardy Jobs “i”)
TABLE I. PROPORTION OF ANNUAL SALES AND PROCESSES AND
MACHINES USED IN THE COSMETIC INDUSTRY IN STUDY. Ti = (Tardiness of Job “i”)
Installed
Capacity
Average Processes for In this model, the setup time is dependent on the
Product % (Total %
Annual Sales their
Type
(2005-2010)
Sales
Manufacture
Available Capability manufacturing sequence.
Processes
Machines)
Lipstick 10.230.000.000 22% 6 11 17%
Mechanic With the following objective function (1):
12.406.250.000 27% 5 11 17%
al Pencils
Wooden
Pencils
15.884.600.000 34% 9 24 38% (1)
Liquids 4.330.466.667 9% 3 5 8%
Viscous 1.999.083.333 4% 4 8 13%
Masks 1.923.825.000 4% 3 5 8% Subject to the following restrictions:
Total 46.774.225.000 100% 30 64 100%
Ti = max (0, Ci - di) ∀_ = 1,… , n

IV. Problem Formulation , (2)

A. Mathematical Model
(3)
In this paper, a mathematical model was proposed for the
Flexible Job Shop Scheduling Problem, whose objective is the (4)
minimization of the number of tardy jobs. The assumptions
were determined, the variables were defined and the (5)
constraints of the mathematical model were established, as
shown below: (6)

The preliminary conditions or assumptions of the model

are as follows:
• The machines are homogeneous per operation, that is to
say, the operating time and production capacity are the
same.
• Setups times are the same for each job on each machine
and are included within the operating time.
• Is known the delivery date of each job.
• Is known the release date of each job.
• Is known what the jobs that can be developed in what
machines.
The following are the model parameters:
i = type of orders (wooden pencils)
j = number of machines
(7)
(8)
The equation (1), represents the objective function, which
seeks to minimize the number of tardy jobs. (2), determines
whether a job is considered to be tardy or not and the way for
calculating the number of tardy jobs. (3), is a precedence
constraint of the same job, that is to say, the total operation
time must be less than or equal to the sum of the setup time
and the operation time. On the other hand (4), shows that the
job cannot start on a machine if it has not finished on the
previous machine. In addition, (5) is a capacity constraint,
taking into account the operations that require the same
machine. (6), states that the same machine cannot have two
jobs at the same time. In (7), the same job cannot be on two

1017 -3-
 Proceedings of 2017 4th International Conference on
Control, Decision and Information Technologies (CoDIT'17) / April 5-7, 2017, Barcelona, Spain

machines at the same time. Finally, (8) establish the conditions

of decision variables.

B. Proposed Algorithm
In the “Fig. 2”, it shows the flowchart for the development
of the genetic algorithm proposed. Then, a computational Fig. 3. Chromosomes father.
pseudocode was developed, for the proposed genetic
algorithm.

Fig. 2. Flowchart of generic pseudocode.

C. Construction of Initial Solution

For construction of the initial solution, two chromosomes
father are generated using heuristics based on dispatching
rules SPT (Shortest Processing Time) and EDD (Earliest Due Fig. 4. Application of the crossover operator PMX.
Date), that basically to built two solutions or sequences in
which the Jobs and the machines taking into account smaller E. Mutation Operator
processing times and nearest delivery dates; these rules were Subsequently, the mutation operator, which allows
chosen taking into account the criterion to be minimized, the intensifying, in the region of possible solutions to the Job
number of Tardy Jobs. According to Holthaus [27], the SPT Shop Scheduling Problem and minimizing the number of
has been found to be the best rule for minimizing the mean Tardy Jobs, which is obtained the optimal solution for the
flow time and the number of tardy jobs. proposed genetic algorithm. Mutation Operator is used by the
exchange. For chromosome that mutates, two chromosomes
D. PMX Crossover Operator that change position is randomly selected, as shown in the
Once generated chromosomes father, as shown in the "Fig. 5" and "Fig. 6".
“Fig. 3”, applies the crossover operator PMX (Partially-
Mapped Crossover), according to Goldberg & Lingle [28],
which allows diversify the range of solutions that meet the
parameters set in the mathematical model, the objective
function and constraints thereof, resulting in a region of
possible solutions that meet the criteria of the model, on which Fig. 5. Chromosome before mutation.
the optimal solution with the mutation operator is sought. The
application of the crossover operator PMX, it shows in the
“Fig. 4”:

1018 -4-
 Proceedings of 2017 4th International Conference on
Control, Decision and Information Technologies (CoDIT'17) / April 5-7, 2017, Barcelona, Spain
Fig. 6. Chromosome after mutation.
V. Experimentation and Analysis
Model Results
A. Factors and Levels for the Factorial Design
According to Gutierrez-Pulido [29], the experimental
design is defined as a statistical methodology for the planning
and analysis of an experiment. Its utility is that they allow
evaluating different situations for a given variable, providing
statistically valid and meaningful information, while an
adequate management of resources is carried out. For this
particular study, the design proposed is a factorial design,
according to McGuigan [30] is defined as: "a design in which
all possible combinations of the selected values of the
independent variables are used". The experimental design
presented is a factorial design 33, that is, of three factors and
three levels of each factor as shown in Table II. The response
variable is the number of tardy jobs corresponding to the
objective function that is pretending minimize.
TABLE II. PROPORTION OF ANNUAL SALES AND PROCESSES AND
MACHINES USED IN THE COSMETIC INDUSTRY IN STUDY.
Factor Notation Values
Probability of Mutation Pm 0,05 0,1
Number of Generations Ngen 100 150
Number of Individuals in the
Nind 10 15
population
To carry out the experiment 10 replicates were used for a
total of 30 data per treatment and are considering 9 treatments
for each factor, has a total of 90 data level factor; then to have
3 factors a total of 270 runs is obtained. For this study, the
SPSS® statistical software was used. The results of the
experimental analysis performed showed that the best
combination of values for the different parameters of the
algorithm are respectively: Pm=0.1, Ngen=200 y Nind =15.
B. Comparative Analysis of Genetic Algorithm Proposed vs.
Company’s schedule of the Company
With the results design factorial experiment, it made a
sensitivity analysis for both the current company’s schedule of
the company and the proposed genetic algorithm, in which
was obtained that the value of the objective function (number
of Tardy Jobs) with variation in the number of production
orders. Was calculated The percentage difference or
percentage relative error between the two methods using the
equation (10). It was shown that in 56% of the runs, reductions
between 20% and 50% were obtained in the number of tardy
jobs, with the genetic algorithm, compared to the company’s
schedule method applied by the company. In Table III (see
Annexed 1), shows the comparison between the proposed
genetic algorithm and the company’s schedule:
−
= × 100
(9)
C. Comparative Analysis Algorithm Proposed vs. Instances of
Literature by Chiang and Fu (2008)
A second analysis was performed by comparing the
methodology proposed in this paper, with the study developed
by Chiang and Fu [13], which combines heuristic approaches
dispatch rules, called a genetic algorithm GA-I-CR, for being
the closest thing to the research done in this approach. The
minimum value obtained by the GA-I-CR algorithm, Chiang
and Fu [13], versus the different values obtained in the runs
made for the genetic algorithm, proposed and then compared
the percentage relative error was calculated, using the equation
(11). The comparison between the two methods is shown in
Table 4 (see Annexed 2).
− −−
= × 100 (10)
−−
These results show that the proposed algorithm has a good
performance with respect to the method proposed by Chiang
0,2 and Fu (2008) because in several of the runs performed the
values obtained by the proposed algorithm, they matched to
200
the methodology of these authors, reason why their Percentage
20 difference is zero. This is a good indicator of the efficiency of
the proposed method, as well as the efficiency, because this
algorithm has a relatively low computation time for the
different evaluated instances, although greater than the GA-I-
CR methodology, but Not in a considerable way.
VI. Conclusions and Future Works
In this work, a genetic algorithm is proposed for the FJSSP
to minimize the number of tardy jobs, taking into account the
characterization of the cosmetic manufacturing process
through the development of a mathematical model.
Computational experiments were carried out using datasets
from the literature. Results showed that very good solutions
can be found using our GA algorithm in comparison with
these instances and Company’s schedule with a considerably
less time and computational effort.
The genetic algorithm proposed exceeds up to 50%
company’s schedule and in the worst cases it equals, but never
below this. The comparison with the method proposed by
Chiang and Fu [13], had a very good performance since in
several of the runs carried the values obtained by the proposed
algorithm, equaled the methodology of these authors, so the
percentage difference is zero. For sensitivity analysis
conducted for varying 50, 65 and 80 production orders, it was
found that improved company’s schedule more than half of
1019 -5-
 Proceedings of 2017 4th International Conference on
Control, Decision and Information Technologies (CoDIT'17) / April 5-7, 2017, Barcelona, Spain
time. On the other hand, the following bullet points resume the
future research that can be undertaken in additional works:
• Extending this approach to problems mono-objective to
multi-objective.
• Set the tool developed to make it more general so that it
can be used in other situations or different production
systems to the job such as open shop.
• Using this methodology with other metaheuristics such
as ant colony, simulated annealing or tabu search,
which work well with genetic algorithms.
• Evaluate alternative hybrid approaches with other
methodologies such as neural networks, fuzzy logic or
Petri nets.
• Further research to other cosmetic companies in the
country.
References
[1] J. Chena, W. Wenga, G. Rongb and S. Fujimura, “Integrating Genetic
Algorithm with Time Control for Just-In-Time Scheduling Problems”.
IFAC-Papers On Line 48-3, 2015, pp. 893–897.
[2] M. Ortiz, D. Neira, G Jimenez and H. Hernandez, “Solving Flexible Job-
Shop Scheduling Problem with Transfer Batches, Setup Times and
Multiple Resources in Apparel Industry”, ICSI 2016, Part II, LNCS
9713, 2016, pp. 47–58.
[3] M. Herrera, “Programación de la Producción: una perspectiva de
productividad y competitividad”, Revista Virtual Pro. Nº 111, 2011.
[4] S. Noor, M. IkramUllah Lali, M. Saqib Nawaz, “Solving Job Shop
Scheduling Problem with Genetic Algorithm”, Sci.Int.(Lahore), 27(4),
2015, pp. 3367-3371
[5] A. S. Jain, S. Meeran. “A State of the Art Review of Job Shop
Scheduling Techniques”, University of Dundee, Dundee, Scotland,
1998, pp. 1 – 48.
[6] M. Gonzalez, “Metaheuristics Solutions to Job Shop Scheduling
Problem with Sequence-Dependent Setup Times, Oviedo, España, 2011,
pp. 21-26.
[7] J. Osorio, T. Motoa, “Hierarchical Production Planning in the Flexible
Job Shop”, Rev. Fac. Ing. Univ. Antioquia N.° 44, Colombia, 2008, pp.
158-171.
[8] J.S. Neufeld, J.N. Gupta, U. Buscher, “A Comprehensive Review of
Flowshop Group Scheduling Literature”, Comput. Oper. Res. 70, 2015,
pp. 56 –74.
[9] K. Cansen, M. Mastrolilli, R. Solis-Oba, “Approximation Algorithms for
Flexible Job Shop Problems”. Proceedings of Latin America Theoretical
informatics. (LATIN 2000), LNCS 1776, 2000, pp. 68 –77.
[10] D. J. Garey (1976). The complexity of Flowshop and Job Shop
Scheduling. Mathematics of Operations Research, 1976.
[11] A. S. Jain, S. Meeran, “Deterministic Job Shop Scheduling: Past,
Present, and Future”, European Journal of Operational Research, Vol.
113, 1999, pp. 390-434.
[12] Asociacion Nacional de Industriales ANDI - Camara Colombiana de la
Industria Cosmetica y de Aseo, “Documento General”, 2015.
[13] T.-C. Chiang and L.-C. Fu. “A rule-centric memetic algorithm to
minimize the number of tardy jobs in the job shop”. International
Journal of Production Research, 46(24), 2008, pp. 6913-6931.
[14] S. Dedieu, “Una adaptación en el diseño para la estrategia de las plantas
del lote de uso múltiple. Computers & Chemical Engineering”, Vol. 23,
Cap 1, 1999, pp, S15-S18.
[15] M. Cruz, J. Solís, D. Juárez (2000), “Un algoritmo de satis factibilidad
para el problema de Job Shop scheduling”, 2000, pp. 1-11.
[16] V. Peña, L. Zumelzu, “Estado del Arte del Job Shop Scheduling
Problem”, 2006, pp. 1-8.
[17] G. Mailing, “Algoritmos heurísticos y el problema de job shop
scheduling”, Universidad de Buenos Aires, 2003, pp. 1-67.
[18] H. Sevilla, C. Zurita, “Uso de meta heurísticas para la optimización de la
secuencia de producción y la asignación de mano de obra en una
empresa manufacturera”, 2010.
[19] G. Moslehi, M. Mahnam, “A Pareto approach to multi-objective flexible
job-shop scheduling problem using particle swarm optimization and
local search2. International Journal of Production Economics, 129(1),
2011, pp. 14–22
[20] Liang Sun 1, 3, Xiaochun Cheng 2, Yanchun Liang, “Solving Job Shop
Scheduling Problem Using Genetic Algorithm with Penalty Function”,
International Journal of Intelligent Information Processing Volume 1,
Number 2, 2010, pp. 65-77.
[21] V. Sels, K. Craeymeersch, M. Vanhoucke, “A hybrid single and dual
population search procedure for the job shop scheduling problem”,
European Journal of Operational Research, vol. 215, issue 3, 2011, pp.
512-523
[22] Drobouchevitch, I.G.; Strusevich, V.A (2000) Two-stage open shop
scheduling with a bottleneck machine. European Journal of Operational
Research 128 (1), 2001, pp. 159-174.
[23] J. Zhenyuan, L. Xiaohong, Y. Jiangyuan, J. Defeng, “Research on job-
shop scheduling problem based on genetic algorithm”, International
Journal of Production Research, Vol 49, Issue 12, 2011, pp. 3585–3604
[24] Euromonitor International, “Statictics Cosmetic Industry Report”, 2016.
[25] F. Miguel, “Dinámica Comercial y Económica del Sector Cosméticos e
Ingredientes Naturales en Colombia”. Programa de Calidad para el
Sector de Cosméticos, ONUDI, Suiza, 2015, pp. 1-106.
[26] Asociacion Nacional de Industriales ANDI - Camara Colombiana de la
Industria Cosmetica y de Aseo, “Documento General”, 2012.
[27] H. Oliver, R. Chandrasekharan, "Efficient dispatching rules for
scheduling in a job shop," International Journal of Production
Economics, vol. 48(1), 1997, pp. 87-105.
[28] D.E. Goldberg, R. Lingle, “Alleles, Loci and the Traveling Salesman
Problem”, Proceedings of the First International Conference on Genetic
Algorithms and their Applications, 1985, pp. 154-159.
[29] H. Gutierrez, R. de la Vara. Analysis and Design of Experiments, 2nd
ed. Mc Graw-Hill, Chapter 1, Introduction to Experiment Design, 2008,
pp. 2-16.
[30] McGuigan, F.J, “Experimental Psychology. Research Methods”, 6th ed.
Mexico: Prentice-Hall. Appendix B. Writing an experiment, 1996, pp.
313-332.
1020 -6-
Proceedings of 2017 4th International Conference on
Control, Decision and Information Technologies (CoDIT'17) / April 5-7, 2017, Barcelona, Spain

Annexed 1
TABLE III. COMPARATIVE ANALYSIS OF GENETIC ALGORITHM PROPOSED VS. COMPANY’S SCHEDULE

1021 -7-
 Proceedings of 2017 4th International Conference on
Control, Decision and Information Technologies (CoDIT'17) / April 5-7, 2017, Barcelona, Spain

Number of late
Number of late Number of late
Number of late works Method Number of late Number of late
works Method works Method 2:
Number of works Method 1: 2: percentage percentage works Method 1: works Method 2: percentage
1: Tool Programming
the run Tool proposal Programming difference difference Tool proposal Programming difference
proposal (65 Manual (65
(50 orders) Manual (50 (80 orders) Manual (80 orders)
orders) orders)
orders)

Run 1 4 4 0% 4 4 0% 3 5 -40%

Run 2 3 4 -25% 3 4 -25% 5 5 0%

Run 3 3 4 -25% 2 4 -50% 5 5 0%

Run 4 2 4 -50% 3 4 -25% 4 5 -20%

Run 5 4 4 0% 4 4 0% 5 5 0%

Run 6 3 4 -25% 2 4 -50% 5 5 0%

Run 7 4 4 0% 3 4 -25% 4 5 -20%

Run 8 2 4 -50% 4 4 0% 5 5 0%

Run 9 4 4 0% 4 4 0% 3 5 -40%

Run 10 2 4 -50% 2 4 -50% 3 5 -40%

Run 11 3 4 -25% 2 4 -50% 3 5 -40%

Run 12 3 4 -25% 3 4 -25% 4 5 -20%

Run 13 4 4 0% 4 4 0% 4 5 -20%

Run 14 3 4 -25% 4 4 0% 3 5 -40%

Run 15 4 4 0% 4 4 0% 3 5 -40%

Run 16 4 4 0% 2 4 -50% 5 5 0%

Run 17 4 4 0% 2 4 -50% 3 5 -40%

Run 18 4 4 0% 4 4 0% 5 5 0%

Run 19 4 4 0% 4 4 0% 4 5 -20%

Run 20 2 4 -50% 4 4 0% 3 5 -40%

Run 21 3 4 -25% 2 4 -50% 3 5 -40%

Run 22 4 4 0% 3 4 -25% 5 5 0%

Run 23 2 4 -50% 4 4 0% 5 5 0%

Run 24 3 4 -25% 3 4 -25% 5 5 0%

Run 25 4 4 0% 3 4 -25% 4 5 -20%

Run 26 3 4 -25% 3 4 -25% 5 5 0%

Run 27 2 4 -50% 2 4 -50% 4 5 -20%

Run 28 4 4 0% 4 4 0% 5 5 0%

Run 29 3 4 -25% 2 4 -50% 3 5 -40%

Run 30 4 4 0% 2 4 -50% 4 5 -20%

Annexed 2
TABLE IV. COMPARATIVE ANALYSIS OF GENETIC ALGORITHM PROPOSED VS. INSTANCES OF LITERATURE BY CHIANG AND FU

1022 -8-
Proceedings of 2017 4th International Conference on
Control, Decision and Information Technologies (CoDIT'17) / April 5-7, 2017, Barcelona, Spain

MINIMUM VALUE LATE WORK

OF NUMBER OF NUMBER VALUE OF NUMBER OF PERCENTAGE
INSTANCE
LATE WORK GA-I - PROPOSED REPS DIFERENCE
CR ALGORITHM

6 7 0%

7 8 17%
ta31 6
8 9 33%

9 6 50%

TOTAL 30 100%

8 6 0%

9 6 13%

ta32 8 10 5 25%

11 8 38%

12 5 50%

TOTAL 30 125%

4 11 0%

ta33 4 5 10 25%

6 9 50%

TOTAL 30 75%

7 10 0%

8 3 14%
ta34 7
9 6 29%

10 11 43%

TOTAL 30 86%

11 7 0%

12 11 9%
ta35 11
13 6 18%

14 6 27%

TOTAL 30 55%

1023 -9-
Proceedings of 2017 4th International Conference on
Control, Decision and Information Technologies (CoDIT'17) / April 5-7, 2017, Barcelona, Spain

MINIMUM VALUE LATE WORK

OF NUMBER OF NUMBER VALUE OF NUMBER OF PERCENTAGE
INSTANCE
LATE WORK GA-I - PROPOSED REPS DIFERENCE
CR ALGORITHM

9 4 0%

10 7 11%
ta36 9
11 9 22%

12 10 33%

TOTAL 30 67%

12 9 0%

13 7 8%
ta37 12
14 7 17%

15 7 25%

TOTAL 30 50%

5 9 0%

6 6 20%
ta38 5
7 8 40%

8 7 60%

TOTAL 30 120%

10 5 0%

11 8 10%
ta39 10
12 7 20%

13 10 30%

TOTAL 30 60%

8 11 0%

9 5 13%
ta40 8
10 9 25%

11 5 38%

TOTAL 30 75%

1024 -10-