http://www.diva-portal.

org

This is the published version of a paper presented at 7th Swedish Production Symposium, Lund,
Sweden, October 25-27, 2016.

Citation for the original published paper:

Ruiz Zúñiga, E., Urenda Moris, M., Syberfeldt, A. (2016)

A simulation-based multi-objective optimization approach for production and logistics
considering the production layout.
In:

N.B. When citing this work, cite the original published paper.

Permanent link to this version:

http://urn.kb.se/resolve?urn=urn:nbn:se:his:diva-13085
 A simulation-based multi-objective optimization approach for production and logistics
considering the production layout
Enrique Ruiz Zúñiga, Matias Urenda Moris, Anna Syberfeldt
Production and Automation Engineering Division
School of Engineering Science, University of Skövde
Högskolevägen 54128, Skövde. Sweden
enrique.ruiz.zuniga@his.se

Abstract
Manufacturing sectors in Sweden have a long tradition and represent a significant share of the national
gross domestic product and the export values. Most of the Swedish manufacturing companies have gone
through a modernization and adaptation process in order to be able to compete on a globalized market.
Many plants, however, still have non-optimized shop floors as a consequence of the shop floors being
adapted over time without redesigning its production and logistics flows and with a lack of an overall
strategy. To support the optimization of shop floors, this paper suggests the combined use of Discrete-
Event Simulation and Simulation-Based Multi-objective Optimization. The aim of the paper is to analyze
a simulation methodology that supports the optimization of shop floors by considering production and
logistics flows along with the shop floor layout. The methodology is intended to contribute to
significantly increase the productivity and efficiency of the Swedish manufacturing industry and help
companies to survive on the globalized market. Through a case study, the paper shows that the proposed
methodology is useful in practice and that it provides a decision support system for manufacturing
companies.
Keywords: Simulation-based optimization, mix-model production, logistics, layout.

methodology can contribute significantly to increase the

1. Introduction productivity and efficiency of the Swedish manufacturing
Competition in the nowadays globalized world is something industry and help them survive on a global market.
really common in the manufacturing sector. There are many The rest of this paper is organized as follows: in Section 2
international companies trying to conquer a bigger share of a literature review of related papers in DES, SBO and
the market and usually a key factor in this task is the manufacturing systems (layout, production and logistics) with
efficiency in the manufacturing processes. A great effort is high product mix and low volume production is presented.
required to achieve the level of efficiency needed to stay Section 3 presents the main steps of the simulation
present in a significant share of the market. This effort can be methodology and how it has been applied to a real-world
supported by new improvement approaches as Discrete-Event problem from the manufacturing industry. In Section 4 initial
Simulation (DES) and multi-objective Simulation-Based results are presented and finally, Section 5 presents
Optimization (SBO). conclusions and future work.
Simulation is an analytical tool to design, maintain and 2. Literature review
improve different types of complex systems [1]. It has been
successfully applied in many sectors by building simulation In this chapter the related scientific literature to this paper is
models that represent accurately different kinds of processes, analysed. The chapter is divided in three main sub-chapters:
systems and even organizations. With the use of DES models, discrete-event simulation, simulation-based multi-objective
system analysis and comparisons of different parts of the optimization and high product mix and low-volume
complex systems can be done easily with a relatively minor production manufacturing system.
investment and without interrupting the production [2].
In this paper an initial attempt to find a methodology for 2.1. Discrete-Event Simulation
the improvement of complex manufacturing systems with Nowadays there are different techniques for process
SBO considering the production and logistics flows and plant improvement. All of them have different applications for a
layout is presented. The research is conducted in close big variety of more specific purposes such us different
collaboration with a middle-sized Swedish manufacturer, process improvement and design or feasibility studies. Some
Xylem Water Solutions. Xylem manufactures water pumps examples of process improvement approaches are linear
and has its main factory in Emmaboda, in the southern region programming, Markov Chain Analysis, DES, System
of Småland. Dynamics, Mathematical Modelling, Montecarlo Simulation
In the paper, a methodology to build conceptual and or Value Stream Mapping and some other Lean approaches.
detailed simulation models of production lines and logistics Simulation techniques can be a suitable approach in diverse
systems is introduced and initial results are presented. It is applications for process improvement of complex systems
demonstrated that the proposed methodology that combines with high variability [3]. This characteristic results really
DES and SBO together can be really useful as a decision useful to evaluate changes or variations in processes,
support system in manufacturing companies. The facilities or procedures in order to increase its efficiency in a
simpler and more effective manner. Another important
benefit of simulation is that it helps to identify the
bottlenecks, the main limitations of the system.
Some authors study the simulation and optimization of
production and logistics systems using mathematical
modelling, such as Petri nets [4, 5]. Due to the high degree of
complexity, variability and stochastic behaviour of the
systems in consideration, this project is mainly focused on
simulation and optimization with DES modelling.
Nevertheless, some mathematical modelling is addressed in
order to support the DES models, for example modelling the
required amount and scheduling of some of the automated
transports that feed the production lines.
“Whether done by hand or on a computer, simulation
involves the generation of an artificial history of a system and
the observation of that artificial system to draw inferences
concerning the operating characteristics of the real system”
[6]. During the 21st century simulation started to be a key
technology to support and improve many different kinds of
systems. Hence, simulation presents a huge potential for
product and manufacturing process development and
improvement [7].
Each time the availability of special-purpose simulation
languages, the massive computing capabilities at a decreasing
cost per operation and the advances in simulation
methodologies make simulation one of the most widely
applied and accepted tools in operation research and system
analysis [8]. Simulation models are usually analyzed by
numerical methods instead of applying analytical ones;
“analytical methods employ the deductive reasoning of
mathematics to “solve” the model; numerical methods
employ computational procedures to “solve” mathematical
models” [8].
The variability and difficulty of the processes within
complex systems usually demand the analytic power of DES.
A discrete-event system is one in which the state of the
system changes at only individual, but possibly random, set
of time points, known as event times [6]. An event is a
change in system state. A simulated clock, provided by the
simulation software, records the time points at which events
occur on a discrete-event simulation [6]. This fact makes Fig. 1: Simulation steps, Banks 2005[8]
DES a suitable tool for the analysis, modelling and
improvement of complex manufacturing systems at discrete As it is possible to appreciate in the previous flowchart, the
events of time. first step when building a simulation model is to have a clear
One of the more highlighted characteristics of DES studies problem formulation. The problem should be perfectly
is the possibility to apply “what-if” questions or scenarios to understood by the client and the simulation analyst. After
the existing systems without disturbing them. Within these this, it is possible to start with the setting of objectives and
scenarios new alternatives, ideas, systems and work overall project plan [9]. The setting of objectives should be
proceedings can be tried out without disturbing the real the answers that will be obtained by implementing a specific
system, or be developed even before a system is constructed. simulation project. Following this previous step, if simulation
When the complexity and variability of the system is is the suitable approach for developing the project, an overall
increased, DES is often necessary in order to be able to project plan has to be done in order to study how the project
model and represent the complex and stochastic flows in will be developed and implemented depending on the
production lines and logistics systems. available resources and time. It should be also considered if
As shown in Figure 1, a DES simulation project is usually the model will be used in the future for other specific studies
compounded by several steps. These steps have to be in the same system; thus, maybe other processes or resources
carefully performed during the building process of the should be taken into account when developing the model.
simulation model in order to obtain accurate models Once the problem, objectives and project plan are clear, it
providing accurate results. is possible to start with the next steps: model
conceptualization and data collection. These both steps are
the “key steps” to perform a good model in order to obtain
accurate results. “The art of modelling is enhanced by an
ability to abstract the essential features of a problem, to select
and modify basic assumptions that characterize the system, material handling systems [7, 11-13]. Nevertheless, there are
and then enrich and elaborate the model until a useful not many cases in the literature where DES and SMO applied
approximation results” [8]. The data collection step is usually together to support the design and development of production
one of the tasks which need more time to be performed. Time and logistics systems of high product mix and low volume
studies may be needed and time standards applied in different production.
processes to obtain accurate data from the real system.
When the required data is collected and the conceptual 2.3. High product mix and low-volume production
model of the system to study it is clear, it is possible to start manufacturing systems
with the model construction. This step means translating the In this subchapter the related literature to this kind of high
conceptual model into a computer language with the help of a product mix and low-volume production regarding the layout,
simulation software tool. This model has to represent all the production lines and logistics is addressed. Common
key factors that are considered for the improvement problems that characterize this kind of systems are the large
approach. Then when the models are built and working amount of manual processes and process steps, the high
properly they have to be verified and validated. Verification variability of process times, the buffers and space constrains,
and validation are two processes to ensure a simulation model the needed specialized resources and the amount of external
represents the system in an accurate manner. Verifying the suppliers. In most of the analyzed papers some of the
models means to check that everything represented in the problems mentioned above are addressed specifically,
model is done properly and validation is to check that the sometimes analyzed from an overall perspective but few of
behavior of the model represents the reality in an accurate them including the layout, logistics and production lines of
way. these high product low-volume production systems. Most of
Once the model is verified and validated it can be analyzed the relevant papers analyzed are related to the automotive
and possible weaknesses of the system and potential sector. Due to the specific case study of this paper, it is not
improvement areas can be found. Finally, when the obtained easy to find related papers on similar production systems,
results are satisfactory they can be presented to decision hence the ones found related to the automotive sector have
makers and stakeholders in order to serve as a decision been utilized as a guide due to the similarities in high product
support system. mix and low-volume production.
2.2. Simulation-based multi-objective optimization Different automation, production and logistics methods in
the manufacturing sector, their main characteristics and
DES provides the results of specific what-if experiment limitations have been deeply analyzed by Groover [14]. A
scenarios. That is a really convenient tool when a limited complete study in the automotive sector of a mixed-model
amount of different possibilities of system configurations are assembly line focused in the logistics including an
considered. However, in order to analyse several scenarios, a optimization is presented by Wenping [15]. Battini [12]
large amount of modelling time is usually required and defines an interesting integrated approach to parts and
although an improved scenario can be found, optimal components management to optimize the centralization
solutions are not guaranteed. Since simulation is not an degree of warehouses to minimize storage costs and to
optimization tool by itself, a step that combines simulation choose the right feeding policies. Ende and Boysen [16]
and optimization is required when several possible summarized the problem of locating the logistics areas
combinations of the system want to be analysed. optimally to facilitate just-in-time supply of mixed-model
Traditionally, simulation and optimization have been assembly lines. According to these authors, in modern-day
considered as different approaches in the operational research production systems there is an increasing challenge to feed
domain, but they have developed together and finally the idea mixed-models production lines due to the ever rising product
is to use the great detail of simulation together with the variety. They propose a mathematical model with an exact
ability of optimization to give optimal solutions [10]. dynamic algorithm to discuss the pros and cons of the
Combining optimization and simulation tools allow decision- supermarket-concept.
makers to quickly determine optimal system configurations, The concept of supermarket can be defined as a
even for complex integrated facilities. decentralized in-house logistics area close to the assembly
Depending on the type of problem to analyse, there are lines that serves as immediate storage for parts delivered with
different optimization methods that can be used in a just-in-time approach [16]. [17] provides a framework to
combination with simulation, of which several are presented design the supermarket and feeding system to automotive
by Figueira and Almada-Lobo [10]. Meta-heuristic mixed-model assembly lines. In this case the decision of
optimization is a flexible approach to examine any solution choosing a central kitting area to supply all the production
space and it is characterized by quickly achieving good lines was based on the amount of required place on the shop
quality solutions, therefore it has usually been used in floor. Both possibilities of having the storage areas needed to
combination with DES [10]. Consequently, the integration of feed the lines located in a dispersed manner by every
meta-heuristic optimization together with simulation is production line or having a central storage supermarket area
necessary if the optimum range of solutions for the given were analysed. It was demonstrated that the second option,
input is followed. Moreover, if there are multiple objectives the central area supermarket, would free much more space
to be analysed at the same time then SBO is the correct especially by the production lines and would reduce
approach to be applied. SBO facilitates the search for trade- drastically the traffic around the production lines where
offs between several conflicting objectives [10]. usually operators have to walk around. Many manufactures
Different authors have analysed how simulation combined around the globe are adopting this supermarket concept to
with optimization techniques can improve production or supply parts to the assembly lines.
 In the field of the line feeding problem including the matter experts [19]. Data collection can be a really time
proper in-house transportation method and route as well as consuming process in this kind of projects. The main problem
considering different types of production lines (with different of the data collection procedure has been the lack of some
automation levels) not much literature has been found. One specific data which was not available. When collecting data
of the closest papers involving these three principles is to build a simulation model, it usually happens that some data
presented by Battini [12], who introduces the line feeding are missing, are too expensive to gather or the data are based
problem and analyses different solutions of optimally in a process that does not exist [6].
locating in house logistics areas to facilitate Just-In-Time In this case there was a lot of available data for the new
supply of mixed-model assembly lines but without technologically adapted production lines but on the other
considering different production methods for different lines hand there was an important lack of data in the manual and
neither the shop floor layout. old production lines, for example there was no available
operators’ interruption data, hence it had to be collected
3. Proposed methodology manually by performing times studies for the different
The proposed methodology of this research is based on the processes. Sometimes data estimation of the needed times if
Design and Creation research strategy [18]. The election of they are not available in the system can be defined with the
this research strategy is based on the creation aspect of the help of experts in the matter [19]. In this project the needed
aim of this project, to create a guideline to support the data was obtained by the logging devices of some processors
improvement and optimization of middle size manufacturing when available, by time studies for most of the manual
systems with high product mix and a low-volume of processes and with the help of the operators, expert matters
customized products, addressing different kinds of production and managers in some cases in which the data was not
and logistics systems and considering the configuration of the directly available. Several visits to the facilities were
shop floor layout. The need of a methodical documentation of organized to meet the installations, resources and staff.
the procedures implemented in the different steps of this Documentation and meetings with the personal working with
project reinforces the election of this research strategy. logistics and production were deeply analyzed. The literature
For showing the industrial applicability of the review and state of the art of previous work, research and
methodology, several main steps based on the simulation developments in simulation and optimizations of
methodology by Banks [8] are described with an application manufacturing systems were at the same time studied to be
on real-world problems. In this project 3D models of the applied to build the simulation models.
system to analyze and improve are created to visualize all the In all the cases the data was verified by expert matters and
processes at the factory in Emmaboda in a realistic way, managers to ensure that no errors or non-reliable data was
being able to involve the different staff in the improvement included. Non-reliable data is generated automatically by the
process. In the following sub-chapters the main steps are system when the data of a product or process is not
explained in detail based on some of the developed introduced correctly by the staff or when some activities are
simulation models of the industrial case study: missing or interrupted (for example when failures or lack of
materials occur, the lead time can be much higher than the
3.1. Step 1: Define objectives maximum sum of process times, but failures and lack of
The first step in the methodology is the definition of the materials are modeled by interruptions at the production line,
objectives. In the case study, these have been defined based not at individual processes). Therefore, some boundaries,
previously specified by the expert matters and managers were
on the requested goals and tasks of the project according to
the necessities of the company. The overall goal is to support established in the different processes and classifications of
the development and optimization of the production logistics, products.
including modifications of production flow and shop floor 3.3. Step 3: Conceptual model
layout of the factory in Emmaboda, Sweden. The work
involves, among other things, to find improvement The next step is the construction of simplified process maps
parameters of the production and logistics flows considering to be used as conceptual models of the system. The
the layout in the optimization approach. The effective simulation models will be based on these conceptual models;
working area / volume of the shop floor must be taken into they require important consideration in order to try to avoid
account when optimizing the production flow. Furthermore, possible future changes in the models afterwards. Applying
the company wants to find the optimal combination of changes in the construction of the models at this level is
production and logistics systems for each product family and usually much easier than doing it in later steps.
also which combination of production and logistics systems All the processes represented in the first version of the
that is optimal from an overall perspective regarding the conceptual models were revised by going through the flow of
layout and the throughput. Answering these questions will the system as a product. In the same way, irrelevant processes
support decision makers at Xylem to increase the production for the purpose of this project were rejected from the process
and efficiency of the system. maps represented in the conceptual models. In order to
explain why and which processes were rejected from the
3.2. Step 2: Data collection conceptual models, a list of assumptions was discussed with
One of the most critical tasks to construct accurate models is the stakeholders of the project. Some of them are
how to get high quality data. High quality data is necessary simplifications or considerations made to simplify the
and often time studies and time standards have to be applied. simulation models without compromising the results. Usually
The needed data can be obtained by using historical records, they have a lack of relevance for the purpose of the
work measurement procedures or estimations from subject simulation models. They also contain those cases in which
the flow of products or materials and the activities involved alternatives of this layout were one of the main improvement
in the system were modified or these unnecessary data or possibilities of the main shop floor at Xylem Emmaboda.
processes simplified in order to avoid unnecessary processes This combination of lines is shown in Figure 2.
that would not affect the simulation results and would make
the simulation extremely complex without adding a
considerable accuracy. Some example are the grouping of
variants with similar characteristics and processing times (for
example variants with the only difference of the color in the
painting, a different performance curve or a different
configuration of the cables usually have not impact in
changing the processing times).
When the conceptual models are carefully studied, built
and checked, it is possible to continue with the general
purpose and go to the next step: to build the simulation
models.
3.4. Step 4: Simulation models
In order to be able to support decision makers to increase
the production and efficiency of this factory through
simulation, virtual models of the different parts of the real
system have to be designed and built according to the defined
objectives and considering important requirements. These
requirements can for example be the lack of enough free
space on the shop floor, the excessive amount of traffic inside
the factory and extensive amount of manpower needed to
produce different models. The simulation models should be
based on previously designed conceptual models that reflect
the behavior, steps and processes of the system in a piece of
paper. To construct the simulation models of a system, an
analytical tool as DES is necessary due to the high Fig. 2: Main shop floor. Production lines
complexity and variability of manufacturing processes. When The simulation models built as part of this study are based
building the simulation model, all the relevant necessary on the different production lines the factory has in their main
processes should be represented to understand the system. shop floor represented in Figure 2. An ongoing study of the
In the case study, a main simulation model was constructed old and mostly manual production lines compared with the
for the production and logistics flows of the main shop floor new adapted ones with a higher level of automation is being
at Emmaboda. This model is compound by several 3D carried out. In the same way, the different possibilities to feed
models of the different production lines and their respective those combinations of production lines with different
feeding systems. In these models, statistical distributions and methods are also being analysed with simulation and stated as
probabilities have been used to model the different ongoing work. These logistics methods being considered to
processing times in order to achieve in the simulation model transport the material between a material preparation area and
the same stochastic behavior of the production, staff and the different production lines are forklifts, conveyors,
resources as in the real factory. Automated Guided Vehicles (AGVs) and tow-trains.
This part involves converting the conceptual models into a Stakeholders are wondering about transforming all the old
computer-recognizable format. In this case were FACTS production lines to the new system and about joining some of
Analyzer [6, 11, 20] and Flexsim the simulation software them together or not. In the following Figures 3 and 4 the
tools selected to make the translation to a computer program. detailed simulation models of both production systems are
At the beginning, simple production lines were built with presented:
FACTS. With this more user-friendly software, generic 2D
simulation models are easily built. The models were built
with the simplest flow of pumps and processes to analyze the
level of detail required to obtain accurate results. When the
model was working properly, it was time to use Flexsim to
introduce, when needed, 3D visuals, more detailed processes,
staff and interruptions that are usually involved in the real
system, additional tasks depending on probabilities and the
classification of different models and variants of products. In
this way more generic concepts can be analyzed with FACTS
easily and those cases with a higher level of detail required
were analyzed using Flexsim (with the correspondent
significant increase of programming time). Four main
different kinds of product models in eight different
production lines were considered to model. Different Fig.3: Simulation model of a manual production line
 Figure 3 presents the simulation models built with Flexsim with all the aspects needed to build up new scenarios to
of one of the old production lines where most of the increase the service level of the system.
processes are manual. In this production system the same Before running the simulations to obtain similar results as
operator works with the same pump during the entire in the real system and before starting with the what-if
production line; all the operators know how to perform all the scenarios, it was necessary to ensure that the model was
manufacturing tasks and the material is fed to the line by correctly built and the results were accurate enough. The
forklifts. The following Figure 4 presents another simulation model has to be an accurate representation of the real-world
model, also built with Flexsim, of one of the new adapted system [6]. Developing a validated simulation model
production lines with a higher level of automation, involves three basic entities: the real-world system under
specialised operators, using kitting and an AGV to transport consideration, a theoretical model of the system and a
the material from the kitting area to the production line. computer-based representation of the model, the simulation
program [6].
Verification is a determination of whether the computer
implementation of the conceptual model is correct [6]. It is to
ensure that the computer program works properly
representing the system in the same way as the real one is.
The input parameters and logical structure of the model has
to be correctly represented in the model to validate it [8].
Every process had to be performed in the model properly,
using just the necessary staff, resources and process times for
every variant. In this case, to perform the verification, every
different pump, process and staff of the model was followed
during a few simulations to ensure they performed their tasks
and processes as they should. The number of resources,
personnel, models and variants and schedules was also
revised.
Fig.4: Simulation model of a new adapted production line The next step was to evaluate the model. The length of
each simulation was established in 30 days. It was considered
Another type of simulation model is presented in Figure 5. that running the model for 1 month would be enough to get
This is a simplified model of the same line shown in Figure 4 accurate results. The data of an average month of production
but built with the simulation software FACTS Analyser. As was considered to build the models. A replication analysis
mentioned before less detailed simulation models can be built and the definition of the warm-up period were implemented,
with this software saving a lot of time when high level of needed to avoid the excess of variability of the output of the
customization and detail is not required. model and to know how accurate the obtained results were. In
order to determinate the warm-up period of the model,
different calculations comparing the output data of the
simulation models during the first simulated days were
performed. The warm-up period is the time needed by the
model to get the normal or usual working flow without
depending on the initial conditions. In this case, the warm-up
period was obtained making comparisons about the results of
different simulations. Making comparisons of the first
simulated days with the rest of them, it was possible to
appreciate that the model was warmed-up after the 1 day of
simulation (depending on the line, just a few hours were
needed to fill the work in process material of the production
lines but 1 whole day was chosen to obtain results for the
Fig.5: Simulation conceptual model of a new adapted production of entire days). Hence, to avoid this variability, a
production line warm-up period of 1 day was established.
The replication analysis is done to determinate how many
Once these simulation models were built it was necessary replications of the model would be necessary to obtain
to verify and validate them in order to ensure they represent accurate data. In this case the data of the standard deviation
the real system in an accurate manner. This is explained in and the mean of running the simulation a few times (with
the following sub-chapter. different number of replications) were introduced, trying to
get as an output the smaller possible value of replications.
3.5. Step 5: Verification and validation
Finally, an acceptable deviation of the results fort eh purpose
The developed simulations models have to be verified and of these studies was obtained running 5 replications of 30
validated in order to design and test some improved scenarios days with a 1 day warm-up period. At this point that the
with the aim to improve the system. Verification and model was verified and giving accurate results, it was time
validation are necessary to ensure that the model properly for the validation process.
represents the real system as it is, without any failure and Validation is a determination of whether the conceptual
model can be substituted for the real system for the purposes
of experimentation [6]. It can be one of the most difficult
steps in a simulation project. The simulation models have to achieve. According to the output results, new input
represent the real system in an accurate way. A constant parameters for the simulation models will be generated by the
communication between the modeler and the stakeholders optimization; these parameters will be introduced in the
was established. During few meetings the results of the model as input parameters to see if the current results can be
simulation were presented and discussed in order to calibrate improved. This is an iterative process that goes on until all
the model and to be able to obtain the desired accurate results the search space of the possible solutions has been covered.
which represent the behavior of the real system. These input parameters can for example be the number and
type of transporters to feed the lines, the number of
3.6. Step 6: System analysis processors and operators, the size of the buffers between
Once the models are verified and validated it is possible to operations… Some of the obtained results by the
perform a system analysis. The possible weaknesses and optimization engine integrated in the FACTS Analyzer
limitations of the real system can be found by analyzing the simulation software determined the optimal size of the
results of simulation models, their variability and stochastic different buffers of one of the main new adapted production
behavior. It is done by going through the simulation models, lines.
analyzing the results and trying to find bottlenecks and The further expected optimization results are an optimal
shifting bottlenecks of the processes [21]. These bottlenecks set of solutions, different combinations of the input
create the delays on the flow of the whole system, increasing parameters of the system, that allow to increase its
the waiting times and reducing the throughput of patients. In performance and efficiency, including the amount and type of
this part of the project “what-if” scenarios are proposed different transports required for the internal logistics, the
according the results of analyzing the system to check different kinds of production lines and some of their
possible improvements in the existing configuration of the characteristics and the possible considered layouts of the
factory. main shop floor.
3.7. Step 7: What-if analysis 4. Initial results
The most common use of simulation is to analyse what-if The preliminary results obtained by the followed
scenarios: new ideas, procedures and potential improvements methodology in this project are based on the presented
can be evaluated without disturbing the real system or be industrial case study. More specific results about the
tested even before a system is constructed. Additionally, validation process and scenarios can be found at [22]. More
“what-if” scenarios have been a useful tool to analyse the generically, the main initial results obtained by going through
implementation and evaluate the results of potential this methodology and the introduced case study can be
improvements in the system. divided in: Firstly, the identification of the key parameters of
With the previously presented simulation models, different the shop floor layout to improve the production. These
production methods were analysed, a bottleneck and shifting parameters are obtained during the system understanding and
bottleneck analysis was performed and different what-if analysis of the project as well as during the simulation
scenarios regarding the logistics methods for the line feeding models construction. In this case they are some specific
problem were designed to analyse the different improvement processors in the specific production lines that after
possibilities. Further what-if scenarios of the different performing a bottleneck analysis and shifting bottleneck
logistics possibilities for the different production methods are analysis were found out to limit the capacity of the system in
being performed and are stated as future work. a significant way. These parameters can be used as the
The implementation in the system of new ideas and new measurement indexes and are needed for the design of the
behaviours of the different processes could cost a big amount experiments and input and output values of the simulation
of time, money, resources and even stopping the normal models and optimization process stated as future work.
activity of the factory. Hence, when changes want to be Secondly, after analyzing the system and designing some
applied to a real system without disturbing it, it is not what-if scenarios with different ideas and concepts that
possible to do it without the use of a powerful model which managers and stakeholders had in mind, refined simulation
simulates the real system as it is. models have been built to work with the focus on specific
potential areas of improvement. Different versions of the here
3.8. Step 8: Optimization presented simulation models have been built in other to
Finally, with all the available and analyzed data form the real analyze those found bottlenecks and how could they be
system, the simulation models and the what-if scenarios, the improved in different scenarios. For example there is case
optimization engine can be designed in order to find the best study going on about line A, one of the new adapted ones, on
possible combination of suitable solutions for the different implementing some improvements in the main bottleneck and
systems and scenarios. The optimization engine is usually translating that improved system to the rest of lines. Finally,
integrated in the simulation software or connected to it in a with an iterative process and the results of the optimization
master-slave manner. The key improvement parameters of studies of the different considered alternatives, the best
the systems, defined in the optimization software, are combinations of production and logistics for this kind of
stablished as input parameters of the simulation models, then system and layout can be found. Another case study going on
the optimization engine runs the simulation for an specific has obtained the optimal capacity of the different buffers of
amount of simulated time and analyzes the output results. A the new adapted lines, reducing the work in progress
proper evolutionary algorithm is chosen in the optimization considerably. Some of actual buffers were over dimensioned
engine usually depending on the size of the search space of and some others were limiting the capacity of production at
the optimization and the amount of established objectives to certain points of time.
Conclusions and future work 3. Ramis, F.J., L. Neriz, and J.A. Sepulveda. A
simulator to improve waiting times at a medical
This paper presents a methodology that supports the
optimization of shop floors by considering production and imaging center. in Winter Simulation Conference.
logistics flows along with the shop floor layout. The 2008. Concepción, Chile: Universidad del Bio-
methodology is based on DES and SBO in order to tackle the Bio, Departamento de Ingeniería Industrial.
complexities of real-world manufacturing problems. As far as 4. Macías, E.J. and M.P. de la Parte, Simulation and
the authors are aware, there is not much research in the
optimization of logistic and production systems
related literature that uses these approaches for
simultaneously optimizing production and logistics flows
using discrete and continuous Petri nets.
with consideration to the shop floor layout. Through an Simulation, 2004. 80(3): p. 143-152.
industrial case study, the paper shows that the methodology is 5. Piera, M.À., et al., Optimization of logistic and
useful in practice and that it provides a decision support manufacturing systems through simulation: a
system that aid manufacturing companies in improving their colored Petri net-based methodology. Simulation,
productivity and efficiency; more specific results on this case
2004. 80(3): p. 121-129.
study can be found in [22]. Decision making is not an easy
task when facing complex systems as several configurations 6. Banks, J., Handbook of Simulation: Principles,
and objectives, not seldom internally conflicting, need to be Methodology, Advances, Applications, and
considered. This methodology can support managers and Practice. 1998, New York, USA: John Wiley &
stakeholders in taking hard decisions and the results obtained Sons, Inc.
from developed simulations models can serve as guidelines to
achieve significant improvements, which is critical nowadays 7. Tempelmeiera, H., Practical considerations in the
to stay present in a globalized world. optimization of flow production systems.
The industrial case study shows applicability of this International Journal of Production Research,
methodology for solving a real-world problem found at a 2003: p. 149-170.
middle-sized water pumps manufacturer. Simulation models
of the existing facilities were developed as part of the study 8. Banks, J., et al., Discrete-Event System Simulation.
as well as properly verified and validated. Based on the 2005, Champaign, Illinois, USA: Pearson Prentice.
simulation models, it was possible to perform what-if 9. Simulation, D.-E., Discrete-Event Simulation.
analysis for the improvement of the production and logistics 2011, Industrial Informatics master's degree.
flows at Xylem’s plant considering the layout of the shop
University of Skövde.
floor. By using simulation, new automation systems, models
and production and logistics methods could be tested without 10. Figueira, G. and B. Almada-Lobo, Hybrid
interrupting the real system. simulation-optimization methods: A taxonomy and
In the next step of this work, additional focus will be put discussion. Simulation Modelling Practice and
on the optimization step of the presented methodology Theory, 2014. 46: p. 118-134.
considering different layout configurations. Different
optimization methods will be evaluated and 11. Ng, A., J. Svensson, and M. Urenda, Introducing
recommendations and guidelines on its implementation as a Simulation-based Optimization for Production
decision support system will be addressed. Systems. Design to Industry: the FACTS Game.
2008.
Acknowledgment
12. Battini, D., et al., Design of the optimal feeding
The authors dedicate special acknowledgment to the
policy in an assembly system. International Journal
University of Skövde and to Xylem Water Solutions for the
support to this research project. Special thanks also to the of Production Economics, 2009. 121(1): p. 233-
IPSI Research School and to the Swedish KK Foundation for 254.
supporting a great deal of this project. Finally but not less, to 13. Persson, A., et al., Metamodel-Assisted
all the Xylem employees at the factory in Emmaboda that Simulation-Based Optimization of a Real-World
have been involved in this project; without their support this
Manufacturing Problem. 2007.
project could not have been possible.
14. Groover, M.P., Automation, Production Systems,
References and Computer-integrated Manufacturing. 2008:
1. Spedding, T.A. and G.Q. Sun, Application of Prentice Hall.
discrete event simulation to the activity based 15. Wenping, L. and H. Yuming. Car sequencing in
costing of manufacturing systems. International mixed-model assembly lines from the perspective
Journal of Production Economics, 1999. 58(3): p. of logistics optimisation. in 2008 IEEE
289-301. International Conference on Automation and
2. Robinson, S., Discrete-Event Simulation: From the Logistics. 2008.
Pioneers to the Present, What Next? The Journal 16. Emde, S. and N. Boysen, Optimally locating in-
of the Operational Research Society, 2005. 56(6): house logistics areas to facilitate JIT-supply of
p. 619-629. mixed-model assembly lines. International Journal
 of Production Economics, 2012. 135(1): p. 393-
402.
17. Faccio, M., et al., Design and simulation of
assembly line feeding systems in the automotive
sector using supermarket, kanbans and tow trains:
a general framework. Journal of Management
Control, 2013. 24(2): p. 187-208.
18. Oates, B.J., Researching Information Systems and
Computing. 2006: SAGE Publications.
19. Freivalds, A. and B.W. Niebel, Niebel's Methods,
Standards, and Work Design. 2009, New York:
Mc Graw Hill.
20. Ng, A.H., J. Bernedixen, and P. Leif. What Does
Multi-Objective Optimization Have to Do with
Bottleneck Improvement of Production Systems? in
The 6th International Swedish Production
Symposium 2014, Gothenburg, September 16–
September 18. 2014.
21. Roser, C., M. Nakano, and M. Tanaka, Shifting
Bottleneck Detection. 2002.
22. Ruiz Zúñiga, E., M. Urenda Moris, and A.
Syberfeldt. Production Logistics Design and
Development Support: A Simulation-Based
Optimization Case Study (WIP). in 2016 Summer
Simulation Multi-Conference. Summer Computer
Simulation Conference (SCSC). 2016. The Society
for Modeling and Simulation International.