You are on page 1of 24

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

net/publication/332414123

A Comparison of Three Simulation Methodologies for a Six-Sigma


Manufacturing Project – A Business Case Study

Article in International Journal of Lean Six Sigma · April 2019


DOI: 10.1108/IJLSS-03-2018-0025

CITATION READS

1 490

3 authors:

Ali Ahmed John Page


UNSW Sydney UNSW Sydney
6 PUBLICATIONS 12 CITATIONS 113 PUBLICATIONS 549 CITATIONS

SEE PROFILE SEE PROFILE

J. F. Olsen
UNSW Sydney
66 PUBLICATIONS 372 CITATIONS

SEE PROFILE

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

Adopting Lean Six Sigma to AnyLogic Simulation in a Manufacturing Environment View project

Self-Organising Swarm View project

All content following this page was uploaded by Ali Ahmed on 17 July 2020.

The user has requested enhancement of the downloaded file.


The current issue and full text archive of this journal is available on Emerald Insight at:
https://www.emerald.com/insight/2040-4166.htm

Simulation
A comparison of three simulation methodologies
methodologies for a Lean Six
Sigma manufacturing project –
a business case study 405
Ali Ahmed, John Page and John Olsen Received 19 March 2018
Revised 9 February 2019
Department of Mechanical and Manufacturing Engineering, Accepted 13 April 2019
University of New South Wales, Sydney, Australia

Abstract
Purpose – This paper aims to compare the prognostic and visualisation capabilities of all the three
simulation paradigms to identify their suitability and rigor in eliminating weaknesses and bottlenecks in a
Lean Six Sigma (LSS) project.
Design/methodology/approach – The paper uses an light-emitting diode (LED) factory as a business
case to show the differences and advantages of using three different simulation techniques to solve a
manufacturing problem.
Findings – Even though this is only one business case that shows how system dynamics (SD) can be
effective in a Six Sigma manufacturing project, more examples are needed to validate this hypothesis
within Six Sigma and Lean manufacturing fields. Even though, discrete-events (DE) and agent-based
(AB) models are both equally well suited to develop the manufacturing processes and the choice of what
to use may be arbitrarily dependent on the available software or the preference of the modeller, hybrid
models seem to become extremely powerful. Therefore, more hybrid models need to be constructed
within LSS (especially when a flowchart and a SIPOC ((Suppliers, Inputs, Process, Outputs and
Customers) table are combined to develop a hybrid model) to achieve the most accurate results with
accurate representation of reality.
Originality/value – There are three commonly used simulation techniques, DE, AB and SD, but choosing
the right simulation methodology has often been a challenge.
Keywords System dynamics, Simulations, Agent-based, Discrete-events, Lean six-sigma
Paper type Research paper

1. Introduction
In today’s fast-changing and competitive market, it is highly important for companies to
plan and monitor their internal process efficiency and effectiveness (Dahlgaard and Mi
Dahlgaard-Park, 2006; Fryer et al., 2007). New technologies and business strategies are
bringing new possibilities for eliminating defects in any process, from manufacturing to
transactional, and from product to service. The most common continuous improvement
methods that are implemented in manufacturing companies are among others: lean
manufacturing (Shah and Ward, 2003), lean Six-Sigma (LSS; Pepper and Spedding, 2010),
total quality management (Yusof and Aspinwall, 2000), continuous improvement (CI, also
known as Kaizen) (Ahmed et al., 2015), agile manufacturing (Gunasekaran, 1999; Yusuf
International Journal of Lean Six
et al., 1999) and Six Sigma (Dahlgaard and Mi Dahlgaard-Park, 2006). Even though all these Sigma
methodologies differ in the details, they all have the same overall goals, which are reducing Vol. 11 No. 3, 2020
pp. 405-427
waste, improving control over the processes, adding value to the final customer and © Emerald Publishing Limited
2040-4166
reducing costs caused by human inefficiency and processes’ weaknesses (Stamm et al., DOI 10.1108/IJLSS-03-2018-0025
IJLSS 2009). With Lean Six Sigma, the focus is to reduce variation in processes and to reduce
11,3 waste, combining the implementations of lean manufacturing and Six Sigma into one
powerful methodology (Pepper and Spedding, 2010; De Koning et al., 2006).
Simulations, on the other hand, are tools used in Lean Six Sigma to understand a system,
analyse its characteristics and design new policies or strategies to implement changes with
added value to the customer. When it comes to simulation, there are three main methods
406 (paradigms) that can be used within Lean Six Sigma: system dynamics (SD), agent-based
(AB) and discrete-events (DE). These three simulation methodologies have the purpose of
replicating a system or process that needs to be analysed into a computational model that
can run dynamically in time (Maidstone, 2012). According to Gurney et al. (2001), this
computational model corresponds to a virtual representation of the real system and can be
used to evaluate what-if scenarios, optimise process parameters and understand
particularities of the system. The three simulations (SD, AB and DE) were preferred because
all can be tested under a single, well-established framework – AnyLogicTM (Borshchev et al.,
2002). Besides, as noted by Maidstone (2012), these three simulation models are the most
commonly used in lean manufacturing, and as such, interfaces and feedbacks between SD
and AB or DE becomes very practicable, technically.
When a system or process requires simulations capabilities, the choice of which of the
three models to use has to be selected before the project commencement, and the choice is
generally dependent on a number of factors including the modeller expertise, the available
software or the biased preference of the management team, often without taking into
consideration the nuances of the different methodologies (Law et al., 2007). Depending on the
paradigm used, different levels of detail can be obtained, from the specific behaviour of
single entities to macro views of a full system. It is also important to consider the time it
takes to develop the simulation considering the different tools available for each of the three
paradigms (Law et al., 2007). Hybrid and multi-method techniques are also possible in this
context and the technology to develop these models exists. However, AnyLogicTM is the only
software that currently supports the development of multi-method simulations integrating
all three of the simulation techniques (Emrich et al., 2007; Garifullin et al., 2007).
The primary objective of this paper is to take advantage of a Six Sigma project developed
for a small factory that produces light-emitting diode (LED) lights, taking two independent
sections of the process and connecting them in a computer simulation. The result will be the
development of three independent models using the three simulation modelling techniques
to understand the nuances related to a manufacturing setting for a Six Sigma project. The
ways to understand the benefits and disadvantages of each model will be done by:
 measuring the time needed to build the model;
 comparing the results obtained;
 understanding the difficulties of modelling certain characteristics of the process;
 measuring the details obtained;
 discussing the readability of the model;
 checking if there are differences in the amount of time it takes for the model to show
results; and
 discussing the challenges of validating the model.

There is limited research concerning the combination of simulation paradigms with LSS.
Only a selected number of scholars such as Ahmed et al. (2015), Borshchev et al. (2002),
Escobar (2015), Ferrin et al. (2005) and Garifullin et al. (2007) have attempted to find a
correlation between various simulation paradigms and LSS and the improvement Simulation
opportunities by their amalgamation. In fact, only Ahmed et al. (2015) and Borshchev et al. methodologies
(2002) described the possibility of using AnyLogicTM (hybrid of SD, AB, and DE) within an
LSS framework in a manufacturing environment. On their part, Ahmed et al. (2015)
demonstrated how the LSS improvement process (DMAIC; Define-Measure-Analyse-
Improve-Control ) can be applied in the AnyLogicTM setting to enable validation of a
manufacturing system’s existing state and its subsequent improvement. Conversely,
Borshchev et al. (2002) discussed approaches to modelling and distributed simulations 407
without dwelling much on LSS or manufacturing processes. Rather they also used
AnyLogicTM (hybrid simulation modelling) to create complex interdependencies between
discrete and continuous time behaviours.
Looking at the past studies, it is argued in this current project, the establishment of such
a framework (that combines and compares various simulation paradigms with LSS) would
contribute to research not only in manufacturing but also in other diverse fields, such as
education and healthcare. As noted by Ahmed et al. (2015), the ideas of combining Six Sigma
and simulation models have only been recent, and research in this area is still far from being
exhaustive. This current research intends to contribute to this area of study by exploring the
three simulation paradigms as guided tools to assist a manufacturing firm with the
decisions to successfully implement LSS approach. This study also demonstrates how
AnyLogicTM, with its powerful predictive and visualisation capabilities, can be used to
compare the performances of the three simulation models, SD, AB and DE, within LSS. As
such, this research is first of its kind to provide an in-depth, practical comparison of the said
models within a manufacturing industry, thus providing an avenue for further research in
the same industry, and beyond.

2. Lean Six Sigma model


In the present day, LSS is emerging as one of the most prevalent quality/business
improvement methodologies in both manufacturing and service industries (Alsyouf et al.,
2018). LSS is a blend of Six Sigma Methodology, a continuous improvement framework, and
the Lean approach, a tool meant to enhance the process effectiveness and efficiency (Antony,
2011; Antony et al., 2003). Six Sigma methodology, as recognised today, was developed at
Motorola through the efforts of Bill Smith and Mikael Harry, reliability engineers, in the
1980s (Brady and Allen, 2006). The framework is composed of a set of statistical tools and
methods used specifically to improve the processes by having better control over them
(Pyzdek and Keller, 2014). In Six Sigma, the processes are improved through variability
minimisation, identifying root causes of defects, and a rigorous process analyses using the
five Six Sigma development phases including DMAIC or Define-Measure-Analyse-Design-
Verify (DMADV) (Ansar et al., 2018). DMAIC is used when the process exists but does not
meet the expectations, whereas, as Cronemyr and Danielsson (2013) argued, DMADV is
used when the product or process does not exist, and a new one needs to be developed.
According to Arcidiacono et al. (2012), LSS adopts the DMAIC framework through which
lean tools combined with Six Sigma tools.
In LSS, the processes to be analysed and improved are identified in the define phase,
whereas the present state is evaluated in the measure stage (Ahmed et al., 2015; Ansar et al.,
2018; Arcidiacono et al., 2012). The potential environment is recognised and validated in the
analyse phase and in the improve phase, solution(s) regarding improved performance levels
and remedial measures are addressed. Finally, in the control phase, enhancing the achieved
gains is done using appropriate engineering, analytical and statistical methods (Ahmed
et al., 2015; Arcidiacono et al., 2012; Hong and Goh, 2003). LSS differs from Six Sigma
IJLSS because it adds removal of waste to the equation, focusing not only on the reduction of
11,3 variation and defects but also integrating lean concepts to achieve continuous flow (Michael
et al., 2004). Its goal is to follow these principles by eliminating the eight forms of wastes in
manufacturing that include transportation, inventory, motion, waiting, over-processing,
over-production, defects and under-utilisation of resources (George, 2002). While Six Sigma
is focused on reducing process failures and improving process output by following a
408 problem-solving approach using statistical metrics, LSS is primarily concerned with
elimination of wastes while at the same time improving the in-house workflow. The
fundamental objective of lean transition is to pinpoint value-adding and non-value-adding
processes. According to Laureani and Antony (2019), LSS allows organisations to reduce the
rising costs, increase customer satisfaction, upsurge profit margins, help develop the
workforce and improve efficiency and effectiveness.

3. Simulation paradigms
Simulations have been used for decades to solve problems that are very dynamic and that are
difficult to tackle by using the statistical methods (Bonabeau, 2002; Fowler and Rose, 2004).
The current scenario in the simulation field presents paradigms that allow users to build
simulation models for various domains (Figueredo and Aickelin, 2011). For instance, in a Six
Sigma analysis based on statistical measurements, simulations models have often been used as
crucial tools within the DMAIC and DMADV phases (Cronemyr, 2007). The most valuable
aspect of simulations in manufacturing Six Sigma projects is their capability to intervene the
system that has to be improved using a virtual representation of the processes that take place,
reducing the risks of failing, and removing the necessity of using real production systems to
test hypotheses, as testing in the real system often requires shutting down production or
reducing production rate (Cronemyr, 2007). By developing simulation models, what-if scenarios,
hypotheses and policies can all be tested without affecting the production system (Zagonel
et al., 2004). These models can help finding root-causes of defects by forcing a better
understanding of the overall system, reducing wastes by testing new work procedures in a safe
virtual environment or designing and testing new technologies to calculate costs, financial
outcomes and customer impact (Bonabeau, 2002; Borshchev and Filippov, 2004).
Three simulation techniques are the most popularly used on a business setting within a
Six Sigma project including DE, AB and SD (Figueredo and Aickelin, 2011; Law et al., 2007),
all of which have different characteristics and ways to approach a problem. However, these
are not the only simulation techniques out there such as dynamic systems (Figueredo and
Aickelin, 2011) and Monte-Carlo simulations (Robert, 2004). Dynamic models can simulate
physical systems such as earthquakes, weather or buildings engineering. Expert systems
can simulate in real-time IF-THEN scenarios to represent the expert decision-making rules
in a computational model (Parunak et al., 1998). Monte-Carlo simulations are used within
any simulation method to understand the variation of the outcome based on a set of
stochastic variables (Robert, 2004). Nevertheless, the focus of this study lies exclusively on
AB, DE and SD, the most commonly used simulation techniques. Each of the three
simulation paradigms is defined and discussed in this section.

3.1 System dynamics


SD is a continuous simulation technique developed in the 1950s by Professor Jay Forrester to
help corporate managers improve their understanding of complex industrial processes
(Borshchev and Filippov, 2004). This simulation technique is based on stocks where entities are
accumulated and flows that define the movement from one stock to the other. Fundamentally,
SD is a technique for understanding the nonlinear behaviour of complex systems over a period
of time using time delays, table functions, internal feedback loops, flows and stocks (Breierova, Simulation
1996; Scholl, 2001). SD has been proven to work in industrial settings, nevertheless it is rarely methodologies
used in manufacturing processes and significantly little research has been done on this matter
[only Baines and Harrison (1999), Lin et al. (1998), and Tesfamariam and Lindberg (2005)], and
the field seems to be relatively inactive. Currently, SD is predominantly used in public and
private sectors for policy analysis and design (Radzicki et al., 2008). Nevertheless, SD provides a
lot of potential other than policy analysis or other problem-centric problems; it also lays out a
framework for understanding a system with variables that are difficult to measure such as 409
awareness, comfort and leadership, among others.
Some of these tools are the causal loop diagram that helps understand the causal
relationships between variables, delays and feedback loop effects in a qualitative manner
(Binder et al., 2004), along with group model building, which is an approach that helps in
generating a mental model consensus among many stakeholders and subject matter experts
(Bérard, 2010; Sterman, 2001). Perhaps the most comprehensive work on the integration of
Six Sigma and SD is that of Cardiel-Ortega et al. (2017) that suggests that although SD is an
obscure methodology, it is a useful, unique tool in Six Sigma methodology that can help
solve problems that elude other continuous improvement techniques. In this study, it is
suggested that SD can be incorporated within LSS to complement existing tools and
complete the loop of knowledge.

3.2 Discrete events


DE is probably the most commonly used simulation model in systems improvement and
operational research. Contrary to SD, it is a very active field, and it has been the most
commonly used simulation technique in manufacturing for decades. As its name suggests,
DE models a process as a series of discrete events (Maidstone, 2012). In the manufacturing
and Six Sigma, this simulation tool’s use is based on the use of sequences of discrete events
that happen in time to represent operational systems. It differs from the SD approach mainly
because of its discrete nature compared to the continuous nature of SD (Maidstone, 2012;
Varga, 2001). DE models use a set of modules or blocks that represent an activity or a set of
activities such as services, queues or batching among others. These blocks are connected
with each other generating a sequence of events that need to be performed to get to a final
state (finished product, completed set of services, etc.; Maidstone, 2012). There is a vast
amount of study in this field related to manufacturing processes. For instance, Zapciu and
Cavalieri (2017) developed a theoretical model in a business case study to optimise buffers to
meet production guidelines. Montevechi et al. (2016) pointed out the increase in studies that
show increased use of DE simulations in manufacturing, as well as the project management
principles applied to these DE models. As noted by Al-Aomar et al. (2015), Six Sigma can be
extended by integrating DES into the DMAIC model. Correspondingly, Escobar (2015)
showed how DE, as a Six Sigma tool, can improve production by helping the firm understand
of how and why products can be re-processed and by determining optimal conveyor speeds.

3.3 Agent-based
AB modelling is a methodology that became widespread in the 1990s. Being a relatively new
method, it is often overlooked in favour of the more established simulation models,
especially the DE (Maidstone, 2012). AB simulations invoke the interaction of different types
of entities or group of entities with the environment, or with other entities to develop a set of
decision-making processes based on simple rules to achieve a certain goal (Railsback et al.,
2006; Siebers et al., 2010). AB simulations use state charts and transitions that are connected
in a network where transitions from one state to the other are based on conditions, rates,
IJLSS timeouts, messages, etc. (Siebers et al., 2010). These simulations are used in manufacturing,
11,3 and a vast amount of research has been done using this methodology. For example, Rocha
et al. (2017) used AB modelling in a manufacturing setting with multi-agents to show how
risks are reduced using this method and to meet requirements for customised products.
Similarly, Paolucci and Sacile (2016) highlighted how in manufacturing processes, all
elements such as machines, products, AGVs or employees can be modelled in a cooperative
410 framework as agents that interact with each other. Project selection and prioritisation are
normally challenging management tasks that require allocation of different kinds of human
resources. Jiang et al. (2014) demonstrated AB Agents can be used in a simulation model to
help address these challenges. The researchers also demonstrated that this technique can be
used as an alternative to classical project selection and definition techniques. Using an AB
simulation for this purpose is also useful for prediction and scheduling purposes, noted
Stranjak et al. (2008), which is an important elements in LSS methodology.

4. Case study analysis


The Six Sigma methodology was applied on a factory that produces LED lights as a specific
case to improve performance, reduce defects and waste and understand the full production
and manufacturing processes within. The define phase of Six Sigma expressed the problems
that the manufacturing time is very high; the defect rate does not meet the expectations and
the costs are high and put at risk the profitability of the factory. As production cannot be
stopped to develop new ideas quickly, using a multi-method simulation model that
incorporates SD, AB and DE in an integrated computational environment was the tool of
choice to analyse the root causes and define the policies to achieve the set goals. As
Figueredo and Aickelin (2011) noted, the choice of one simulation method over the other is
normally not straightforward and is based on subjective matters such as the expertise of the
modeller, the available software or the biased preferences of the management team.
With the problem already identified in this study, two tools (tasks to be developed) were
considered important in comparing the three simulations within DMAIC framework:
(1) For the first task, a flowchart was developed, and three independent simulation
models were built on the basis of that flowchart with three simulation techniques,
expecting to reveal reasons why one simulation methodology may be better than
the others. The flowchart helped in explaining the current state of the LED
manufacturing process.
(2) The second task necessitated the development of a SIPOC table (Suppliers, Inputs,
Process, Outputs and Customers) giving a high-level overview of the task and three
independent simulations models built again using the three simulation techniques.
Fundamentally, the SIPOC table aided in explaining the external manufacturing processes
carried out outside the factory. These two tasks were later linked together to have a full
working model. The research is expected to reveal what is the best method to transform a
manufacturing flowchart into a simulation model and what is the best model to transform
a SIPOC table into a simulation model.

4.1 Potting process flowchart


Flowcharts are widely used by Six Sigma professionals during the measure phase to provide
an overview of the different actors, interactions and lead times that exist in different processes
(Tolga Taner et al., 2007). A flowchart defines a process and allows stakeholders to have a clear
view of the organisational behaviour to enable decision-making by defining new steps,
reallocating resources or optimising the process in question (Arnheiter and Maleyeff, 2005). A Simulation
flowchart is a simple graphical tool that documents the flow of a process, and it is shown methodologies
through the business process modelling notation or business process modeling notation
(BPMN) (Pyzdek and Keller, 2014). The production of LED lights is a process that combines
heat sinks, reflectors and printed circuit board into one product. This product (called heat sink
for this study), when combined, goes through a potting process. During the Six Sigma measure
phase, a flowchart was constructed to provide an overview of different actors and interactions
in the potting process. Figure 1 shows the schematic flowchart demonstrating the process used
411
to maintain continuous improvement in process production.
Heat sinks arrive at an average rate of 3.5 per hour. An employee will start the potting
process by mixing the RTV615 silicon rubber compound and inserting it into each lens pocket
to cover only the LED light; this requires one full-time resource that takes 6 min to complete the
task. After that, the heat sink is placed in a heating room with a capacity of 140 products for
6 h. No resources are needed during the heating process. Afterwards, the product goes to a new
potting procedure where the lens pockets are fully covered with the RTV615 silicon rubber
compound. An employee performs this procedure, which takes 18 min to complete. This
employee will then place the heat sink into the heating room again. Once done, the unit will be
tested, and if rejected, it has to go to the potting procedure again. The unit testing process also
requires the full commitment of an employee who will take 5 min to complete the task and will
reject it 15 per cent of the time. A successful product will move forward along the production
line. This flowchart is the graphical tool that will be the basis for the development of the first
segment of the three independent models. All three models will be validated on the basis of the
products entering the system, the resulting production level and the workforce utilisation.

4.2 SIPOC diagram


In process improvement, SIPOC is a Six Sigma tool that shows the inputs and outputs of a
process in a very clear format to show a high-level perspective to people that are not
necessarily familiar with the details of the process (Pyzdek, and Keller, 2014). Suppliers and
customers can be internal, and the main focus differs from a flowchart in that it does not try
to understand the granularity of a process, but it tries to find (and show clearly) the inputs
and outputs associated (Simon, 2007). In that sense, a SIPOC diagram can be a black box
with a very high-level definition of the process. During the LED production process, there is
an external activity that supplies breath inserts to the factory to combine them with the heat
sinks and build the final product that will then be packaged and distributed to the final
customers. The breath inserts development, being an external process, is not part of the
factory that produces LED lights, but a high-level overview is necessary to understand the
outputs to integrate them into the simulation model.

Figure 1.
Potting process
flowchart
IJLSS Figure 2 shows the SIPOC table for the breath inserts process. Parts 1 and 2 are both used
11,3 together to generate breath inserts that will be combined then with heat sinks; both parts
have the same inputs. The critical volume is equal to 100, and it represents the minimum
parts that need to be available before having new parts delivered from the suppliers. The
arrival quantity of 250 defines the average number of parts that are ordered each time an
order is made. Storage size of 1,250 defines the maximum number of parts that can be
412 stored. The breath inserts capacity of 10 is the maximum number of breath inserts that need
to be stored to deliver them when the LED factory requires them. The outputs from the
SIPOC model are the breath inserts that will be delivered when required to the factory and
the package size, which is the number of breath inserts that are needed to create one unique
heat sink final product. The customers are the LED factory (and its employees) because it
needs the breath inserts to have a good flow in the processes.
This SIPOC table will be the basis for the development of the second part of the
simulation. It differs from the first part that uses a flowchart in the sense that it does not
show very specific details on the process itself but defines the inputs and outputs with high-
level information on what happens. This part is important to represent the full system and is
expected to reveal different needs from a simulation standpoint. The LED factory requires
having an inventory of eight breath inserts at all times and the model will be based on that
fact to have the breath inserts delivered appropriately.

4.3 Simulation models within potting and breath insert processes


4.3.1 Discrete event model. The DE model for this study was built using AnyLogicTM and
the potting process flowchart based on Figure 1 is shown in Figure 3. The blocks used to
build the simulation of the potting process are described in Figure 4.
The model was simulated for 1,000 h starting from zero having an average of 836.5 heat
sinks arrivals during this time and 95 per cent of the heat sinks processed, as shown in
Figure 5. The workforce utilisation results in an average of 91 per cent for two workers and
63 per cent for three workers.

Figure 2.
Breath inserts SIPOC
diagram
Simulation
methodologies

413
Figure 3.
Discrete events model
potting process

Figure 4.
AnyLogicTM DE
building blocks

Figure 5.
Accumulated heat
sinks arrivals and
processed/1,000 h in
DE simulation
IJLSS The second part of the model defines the external process and the combination of breath
11,3 inserts with heat sinks. Figure 6 shows look of the module and Figure7s hows the additional
modules needed to complete the model.
The external process controls the number of breath inserts available in the BIDelivered
queue module. Whenever the number of breath inserts delivered is less than 8, new breath
inserts are delivered and if inventory is low, Parts 1 and 2 are requested to have it in
414 inventory to produce more breath inserts. Figure 8 shows the number of breath inserts
available during a 1,000-h simulation. From Figure 8, it is possible to see that the DE model
can control the number of available breath inserts while maintaining generally eight in
inventory.
4.3.2 Agent-based model. The AB model was also developed using AnyLogicTM. The AB
toolset is less diverse than the DE toolset and is composed by events that run cyclically as
well as java functions, state charts and dynamic variables. The elements used in this
simulation are identified and described in Figure 9.
For the potting process, two specific agents were created: the worker and the heat sink.
Figure 10 shows the main view where the parameters and agents are defined and the arrival
of heat sinks to the system are determined by an event function with the same rate as the

Figure 6.
Discrete event model
breath inserts process

Figure 7.
Additional
AnyLogicTM DE
building blocks
source block in the DE model (3.5 per h). Besides that, two variables hold information on the Simulation
heat sinks that are out of the process and the ones that are in the storage room, whereas the methodologies
utilisation variable holds the information on the workers that are busy.
Figure 11 shows the workforce agent defined by three unique states: idle, busy or off
schedule. This agent has a function that checks the existent heat sinks to find the one
waiting to be managed.
Figure 12 shows the state chart of the heat sink agent. This agent follows the same path
as the one defined in the DE model with a potting 5 per cent (fivePotting) with a queue 415

Figure 8.
Breath inserts
available in DE
simulation

Figure 9.
Elements used in
agent-based
simulation

Figure 10.
AB main agent
potting process
IJLSS
11,3

416

Figure 11.
AB work force agent
potting process

Figure 12.
AB heat sink agent
potting process

(waitingForWorker). Analogically, all states that define a service or process have a waiting
queue where the agent waits for an available agent to be idle. A function is also defined to
check for available workers to process this heat sink whenever it is waiting to be processed.
The model was simulated for 1,000 h starting empty and having an average of 836.5 heat
sinks arrivals during this time, and 95 per cent of the heat sinks processed as shown in
Figure 13. The workforce utilisation results in an average of 91 per cent for two workers and
63 per cent for three workers, which are the same results as in the DE simulation (Figure 5).
This validates both models empirically, as observed Maidstone (2012), results-wise
supporting the fact that both modelling techniques are equivalent in solving this unique
problem. The second part of the model (external breath inserts process) is shown in
Figure 14. It defines external breath inserts process and the combination of breath inserts
with heat sinks. Figure 14 shows how the module looks like and Figure 9 shows the Simulation
additional modules needed to complete the model. methodologies
Figure 15 shows the number of breath inserts available during a 1,000 hours simulation.
From this figure, it is possible to see that the AB model can also control the number of
available breath inserts maintaining generally eight in inventory confirming that DE and
AB simulations are equivalent in solving this problem.
4.3.3 System dynamics model. Like DE and AB, the SD model was developed using
AnyLogicTM, and its toolset is composed using flows, stocks and dynamic variables 417
(Figure 16). All these elements need to be explicitly connected with each other through links
that define their relationships.

Figure 13.
AB Accumulated
heat sinks arrivals
and processed/1,000 h

Figure 14.
AB Part1, Part 2 and
breath insert state
charts and additional
heat sink states

Figure 15.
AB simulation
breath inserts
IJLSS The structure of the SD model is similar to that of DE model and the heat sink agent in the
11,3 AB model. But in SD, it is not possible to explicitly have relationships between services or
tasks and human resources that perform those tasks, as it is a continuous simulation
technique and elements in the model will be connected as opposed to discrete numbers. This
means that, in any of the processes defined in the flowchart, it is possible to have a fraction
of heat sinks such as 0.4 or 1.3. In this case, if there is a workforce of two employees, the sum
418 of the stocks that represent the processes that need human resources (potting 5 per cent,
potting 100 per cent and testing units) cannot be greater than 2; meaning that there can be a
fraction of an employee working in a fraction of a heat sink in a certain process. Even
though this is impossible in real life, the average is accurate and represents the reality in an
abstract way. The SD model that represents the complete potting process is shown in
Figure 17.
Like DE and AB, the SD model was simulated for 1,000 hours from zero, having an
average of 836.5 heatsinks arrivals during this time and 95 per cent of the heat sinks
processed as shown in Figure 18. The workforce utilisation results in an average of 91 per
cent for two workers and 63 per cent for three workers, which are the same results than the
DE and AB simulation. This validates both models and supports the fact that all the three
modelling techniques can be used to solve this problem.
Using AnyLogicTM, the second part of the model that depicts the external breath inserts
process was developed, as shown in Figure 19.
Although the three simulation paradigms show consistency in improving the LED
factory system, Figure 20 shows that there is a considerable difference between SD’s breath
inserts and those of DE and AB available during a 1,000-h simulation. The LED factory
requires having an inventory of eight breath inserts at all times and the model will be based

Figure 16.
AnyLogicTM SD
building blocks

Figure 17.
System dynamics
model potting process
on that fact to have the breath inserts delivered appropriately. From Figure 20, it is evident Simulation
that the continuous nature of SD allows better control around eight breath inserts compared methodologies
to Figure 8 using DE and Figure 15 using AB, but less realistic, as it uses fractions of breath
inserts. Interestingly, there is a difference between Figure 8 and Figure 20, as Figure 8 is
based on discrete numbers, whereas Figure 20 is the system dynamics model, which is
continuous, so there can be 0.2 breath inserts down and the system immediately replenishes.
But this is not realistic because 0.2 breath inserts is not possible. Figure 8 is discrete so we
can only replenish when we are 1, 2, 3 items of breath inserts down. This is a considerable 419
difference between DE (agent based is also discrete) and SD.

Figure 18.
SD accumulated heat
sinks arrivals and
processed per 1,000 h

Figure 19.
System dynamics
model breath inserts
process

Figure 20.
System dynamics
breath inserts
IJLSS 5. Results analysis
11,3 The LED factory processes of potting and breath inserts were modelled and then combined
into one unique simulation model. The following metrics were used to understand the
differences in the results:
 the time to build the models;
 the accuracy or variance of the results;
420  the readability;
 the limitations of the models;
 the differences in the detail of the models;
 the processing speed;
 the animation capabilities; and
 the validation and debugging challenges.

5.1 Time used to build simulation models


To build each model, it is necessary to have a good knowledge of the tools and techniques
available and how to use these techniques to build a working model without conceptual or
structural errors. From that perspective, it is also noticeable that a DE model is much more
compact, as its modules are defined explicitly to work on manufacturing processes, whereas
an SD or AB modeller must define these elements from scratch to complete the model.
Finally, the visual size of the model is proportional to the time it takes to build it. Figure 21
shows the relative amount of time it takes, 1 being the minimum time, so a relative time of 2
means that it takes 2 times as much time to build the model.
Figure 21 shows that building the DE model is much faster, as the modules are already
constructed and it took around 15 per cent more time to build the AB model and 250 per cent
more to build the SD model. The reason why an SD model takes much longer to build is that
it is more difficult to find a way to represent a discrete reality on a continuous structure.
While in DE and AB, it is relatively simple to define a resource that works in a particular
process and a maximum number of those resources. In SD, one has to be more creative to
represent that, and the hypothesis that different modellers will provide different models for
the same process is quite reasonable. A hypothesis that should be tested in future studies is
one that analyses the variability of the model representation created by different modellers

Figure 21.
Relative time to build
the simulations
to understand how creative one can be to obtain the same result. It would be expected that a Simulation
DE model for the potting process would be very similar if it were created by different methodologies
modellers, but an SD model may be extremely different. If this hypothesis were true, the
time to be creative to achieve the modelling purpose would add a substantial amount of time
to the development process.

421
5.2 Accuracy of results
Even though different models are different abstractions of a real system, to validate the
models, they should represent the reality appropriately and accurately. In the case of
potting, two employees work doing that labour at full capacity, and the empirical data
shows utilisation of these resources of 90-95 per cent, meaning that the models should be in
that range. To validate all the models, for three available workers, all three models should
have the same results. Figure 22 shows the utilisation with two and three workers. The
variation in the results for each model is very low. This validates the models and shows that
all models can represent this reality in different ways. However, as SD is a continuous based
simulation technique, discretisation is not possible, or at least not a good practice, so in the
model built, it is possible to have fraction of a worker processing a fraction of a product.
Even though this is not possible in real life, it can still throw correct averages, so the model
becomes correct from a macro perspective even though the granularity of representing
products and workers is not accurate.

5.3 Level of detail and capabilities of the modelling techniques


Manufacturing processes are not modelled using system dynamics, but it has been shown
that it is possible to do it obtaining very similar results when measuring high-level drivers.
However, it is not possible to know how many people are working on a particular process in
a particular time, which may or may not be necessarily dependent on the scope of the
project. On the other side, AB is extremely flexible, and there are so many ways of modelling
the same thing and having the same results, that no two modellers will create the same
model even though the results will be the same. DE is optimized for manufacturing and also
has the flexibility to customise processes in any way.

Figure 22.
Resource utilization
IJLSS 5.4 Readability
11,3 In DE paradigm, it is very easy to see the products that are present in each part of the
process and the number of workers that are active, but it is not possible to see which worker
is working in which process. The AB model, on the other hand, lets the user know in a more
granular level which product is being processed in which process by which employee. But it
is difficult to know how many products are being processed unless an artificial variable is
422 created for this purpose. Reading through the simulation becomes hard when there are too
many existing products making it difficult to find the product that needs to be analysed, and
this is an extra difficulty also when it comes to validate and debug the model. The SD model
does not give any information about products or employees, and it just gives a very high-
level vision of how the average production and utilisation is. This can be useful in the breath
inserts process where information about the process starts at a very high level with the
SIPOC diagram and information about the process is not detailed. A general overview of
what happens is more than enough from a simulation standpoint.

5.5 Processing speed


All three models were run, and processing time was calculated for the full model including
breath inserts and potting processes. Figure 23 shows the relative time it took for each
model to run, 1 being the minimum amount of time and a value of 2 means that it takes twice
as much time to run a model to get final results. Figure 23 shows that the DE and AB models
take almost the same time to run, and SD is 1.8 slower using Euler method for differential
equations with a time step of 0.001 and running the simulation for 1,000 virtual hours.

5.6 Model validation and animation


In a manufacturing process, it is important to use animations as one of the subjective
methodologies used to validate a model. DE and AB have similar animation capabilities, and
they both can show how potting is performed and easily add movement from one point to
the other as part of the process. Even though it is also possible to add these delays in SD, it
not possible to use SD to show an animation. As the level of detail is not achievable in SD, it
more difficult to validate.

Figure 23.
Execution speed
comparison
5.7 Findings summary Simulation
In this study, it was established that there are at least three fundamentally different simulation methodologies
paradigms that can be useful for improving the potting and breath insert processes in an LED
factory. DE is probably the most widely used simulation model in lean manufacturing because
it is fast (regarding execution speed and build speed), flexible and produces good outcomes.
The DE models are built using entities (objects that move through the system), events
(processes that entities pass through) and resources (objects that needed to trigger events). AB
simulation model is a relatively new technique in lean manufacturing and operational research 423
but has a performance capability comparable to DE. In this project, the AB simulation
consisted of autonomous Agents (which are the self-directed objects that move about the
system) and rules (which are followed by the agents to achieve the intended objectives).
For this project, DE and AB were shown to produce the same results in solving LED factory
system problem. SD takes a different approach than both DE and AB, focusing on flows
around networks rather than queuing systems. SD in this project, like in lean projects, involved
the stocks (as basic stores of projects), flows (that define the movements of objects between
different stocks) and delays (delays the measuring and the acting on the measurement). From
the case study, it was evident that SD simulation paradigm performs better in terms of
controlling processes. This was mainly because of the continuous nature of SD, but the results
were considered less realistic, as it uses fractions of breath inserts. Figure 24 summarises the
comparison of the three simulation paradigms within an LSS project.

6. Limitations and future work


There are various limitations evident in this study. First, only SD, AB and DE events were
used in this comparative study. In future projects, more deterministic and stochastic
simulation models need to be considered. Second, the study was conducted in a simple LED
factory setting with only two tasks of flowcharts and SIPOC table. Although accurate and
relevant in this area of study, the results and findings cannot be generalisable. Future
studies need to consider bigger, and perhaps more complex industry setting to improve
generalisability and applicability in other industries. Industries such as healthcare, military,
semiconductor manufacturing, logistics, supply chain, and distribution, project
management, construction engineering, network simulation and risk analysis can hugely
benefit from breakthroughs in integration of simulations and Six Sigma.
While this single one business case shows how SD can be effective in a Six Sigma
manufacturing project, more examples are needed to validate this hypothesis within Six Sigma

Figure 24.
Comparison of three
simulation
paradigms within a
lean manufacturing
project
IJLSS and lean manufacturing fields. Furthermore, DE and AB models are both equally well suited to
11,3 develop manufacturing processes and the choice of what to use may be arbitrarily dependent
on the available Software or the preference of the modeller, hybrid models seem to become
extremely powerful. Therefore, more hybrid models need to be constructed within LSS
(especially when a flowchart and a SIPOC table are combined to develop a hybrid model) to
achieve the most accurate results with accurate representation of reality.
424
7. Conclusion
In the past decade, simulation technology has evolved immensely due to increased computer
power, new simulation software in the market, updated software packages with more
functionalities and an increase in demand. The current research in this field presents
paradigms that allow modellers to build simulation models for various domains within Lean
Six Sigma. However, as pointed out previously, there is relatively few research on the
comparison and combination of the three most commonly used simulation paradigms in
the form of DE, AB and SD. Only Ahmed et al. (2015) and Borshchev et al. (2002) illustrated
the scenarios where hybrid simulations including SD, AB and DE can be used alongside an
LSS framework in a manufacturing environment. Nonetheless, a research that compares the
performance, differences and advantages of the three said simulation paradigms in defect
prevention within is completely missing.
This paper aimed to fill this research gap by demonstrating how the three simulations
compare within an LSS project. This research is first of its kind in simulations, lean, and six
sigma industries, and is expected to be a valuable addition in this new area of study.
Accordingly, the primary objective of this project was to take two small independent processes
in an LED factory and then connecting them together in a computer simulation. In doing so, it
bridges the gap between theory and practice by practically developing three independent
models using the three simulation modelling techniques to understand the nuances related to a
manufacturing setting for a Six Sigma project. The three models were then compared to
determine their suitability and performance within a Lean Six Sigma project.
Although all the three simulation paradigms were shown to solve the problem within LED
factory system, there were evident differences regarding execution speed, detail, flexibility,
building time, SIPOC score and outcome. It was also concluded that AnyLogicTM is a good fit for
DMAIC as part of design for all the five stages. The quality of any manufacturing process may
be improved by utilising LSS in AnyLogicTM simulation environment bearing the potential to
offer beneficial approach in terms of reducing bottlenecks, improving cost and reducing costs.
Moreover, the increased need for businesses to reduce waste due to productivity (economic) and
environmental concerns (humanitarian) highlights the importance of this research, both at
present and in the future. However, this research was also limited in several ways in the areas
especially in the areas of methodology and approach. Future studies are, therefore, necessitated
to address the aforementioned limitations with the intention of further understanding the
implications and new insights on the integration of Six Sigma and simulation models.

References
Ahmed, A., Page, J. and Olsen, J. (2015), “Adopting lean six sigma to AnyLogic simulation in a
manufacturing environment”, 21st International Congress on Modelling and Simulation, Gold
Coast, Australia, pp. 29-35.
Al-Aomar, R., Williams, E.J. and Ülgen, O.M. (2015), “Simulation for six sigma”, Process Simulation
Using WITNESS®, John Wiley & Sons, Hoboken, New Jersey, pp. 489-548.
Alsyouf, I., Kumar, U., Al-Ashi, L. and Al-Hammadi, M. (2018), “Improving baggage flow in the Simulation
baggage handling system at a UAE-based airline using lean six sigma tools”, Quality
Engineering, Vol. 30 No. 3, pp. 1-21.
methodologies
Ansar, A.R., Shaju, S.U.C., Sarkar, S.K., Hashem, M.Z., Hasan, S.K. and Islam, U. (2018), “Application of six
sigma using define measure analyze improve control (DMAIC) methodology in garment sector”,
Independent Journal of Management and Production, Vol. 9 No. 3, pp. 810-826.
Antony, J. (2011), “Six sigma vs lean: some perspectives from leading academics and practitioners”,
International Journal of Productivity and Performance Management, Vol. 60 No. 2, pp. 185-190. 425
Antony, J., Escamilla, J.L. and Caine, P. (2003), “Lean sigma [production and supply chain
management]”, Manufacturing Engineer, Vol. 82 No. 2, pp. 40-42.
Arcidiacono, G., Calabrese, C. and Yang, K. (2012), Leading Processes to Lead Companies: Lean Six
Sigma: Kaizen Leader & Green Belt Handbook, Springer Science & Business Media.
Arnheiter, E.D. and Maleyeff, J. (2005), “The integration of lean management and six sigma”, The TQM
Magazine, Vol. 17 No. 1, pp. 5-18.
Baines, T. and Harrison, D. (1999), “An opportunity for system dynamics in manufacturing system
modelling”, Production Planning and Control, Vol. 10 No. 6, pp. 542-552.
Bérard, C. (2010), “Group model building using system dynamics: an analysis of methodological
frameworks”, Electronic Journal of Business Research Methods, Vol. 8 No. 1, pp. 35-45.
Binder, T., Vox, A., Belyazid, S., Haraldsson, H. and Svensson, M. (2004), “Developing system dynamics
models from causal loop diagrams”, Proceedings of the 22nd International Conference of the
System Dynamic Society. pp. 1-21.
Bonabeau, E. (2002), “Agent-based modeling: methods and techniques for simulating human
systems”, Proceedings of the National Academy of Sciences, Vol. 99 No. Supplement 3,
pp. 7280-7287.
Borshchev, A. and Filippov, A. (2004), “From system dynamics and discrete event to practical agent
based modeling: reasons, techniques, tools”, Proceedings of the 22nd International Conference of
The System Dynamics Society, Vol. 22.
Borshchev, A., Karpov, Y. and Kharitonov, V. (2002), “Distributed simulation of hybrid systems with
AnyLogic and HLA”, Future Generation Computer Systems, Vol. 18 No. 6, pp. 829-839.
Brady, J.E. and Allen, T.T. (2006), “Six sigma literature: a review and agenda for future research”,
Quality and Reliability Engineering International, Vol. 22 No. 3, pp. 335-367.
Breierova, L. (1996), “An introduction to sensitivity analysis. MIT system dynamics in education
project”, available at: http://sysdyn.clexchange.org/sdep/Roadmaps/RM8/D-4526-2.pdf
Cardiel-Ortega, J., Baeza-Serrato, R. and Lizarraga-Morales, R. (2017), “Development of a system dynamics
model based on six sigma methodology”, Ingeniería e Investigacion, Vol. 37 No. 1, pp. 80-90.
Cronemyr, P. (2007), “DMAIC and DMADV-differences, similarities and synergies”, International
Journal of Six Sigma and Competitive Advantage, Vol. 3 No. 3, pp. 193-209.
Cronemyr, P. and Danielsson, M. (2013), “Process management 1-2-3 – a maturity model and
diagnostics tool”, Total Quality Management and Business Excellence, Vol. 24 Nos 7/8,
pp. 933-944.
Dahlgaard, J.J. and Mi Dahlgaard-Park, S. (2006), “Lean production, six sigma quality, TQM and
company culture”, The TQM Magazine, Vol. 18 No. 3, pp. 263-281.
De Koning, H., Verver, J.P., van den Heuvel, J., Bisgaard, S. and Does, R.J. (2006), “Lean six sigma in
healthcare”, Journal for Healthcare Quality, Vol. 28 No. 2, pp. 4-11.
Emrich, S., Suslov, S. and Judex, F. (2007), “Fully agent based. Modellings of epidemic spread using
AnyLogicTM”, Proc. EUROSIM, pp. 9-13.
Escobar, C. (2015), “A lean six sigma analysis supported by discrete event simulation for pecan
production improvement”, Paper presented at the IIE Annual Conference. Proceedings, p. 1.
IJLSS Ferrin, D.M., Miller, M.J. and Muthler, D. (2005), “Lean sigma and simulation, so what’s the correlation?
V2”, Simulation Conference, 2005 Proceedings of the Winter, IEEE, Vol. 5.
11,3
Figueredo, G.P. and Aickelin, U. (2011), “Comparing system dynamics and agent-based simulation for
tumour growth and its interactions with effector cells”, Proceedings of the 2011 Summer
Computer Simulation Conference, pp. 52-59.
Fowler, J.W. and Rose, O. (2004), “Grand challenges in modeling and simulation of complex
manufacturing systems”, Simulation, Vol. 80 No. 9, pp. 469-476.
426
Fryer, K.J., Antony, J. and Douglas, A. (2007), “Critical success factors of continuous improvement in
the public sector: a literature review and some key findings”, The TQM Magazine, Vol. 19 No. 5,
pp. 497-517.
Garifullin, M., Borshchev, A. and Popkov, T. (2007), “Using AnyLogicTM and agent-based approach to
model consumer market”, Proceedings of the 6th EUROSIM Congress on Modelling and
Simulation, pp. 1-5.
George, M. (2002), Lean Six Sigma: combining Six Sigma Quality with Lean Production Speed,
Combining Six Sigma Quality with Lean Production Speed, McGraw Hill Professional, PA.
Gunasekaran, A. (1999), “Agile manufacturing: a framework for research and development”,
International Journal of Production Economics, Vol. 62 Nos 1/2, pp. 87-105.
Gurney, K., Prescott, T.J. and Redgrave, P. (2001), “A computational model of action selection in the basal
ganglia. II. Analysis and simulation of behaviour”, Biological Cybernetics, Vol. 84 No. 6, pp. 411-423.
Hong, G.Y. and Goh, T.N. (2003), “Six sigma in software quality”, The TQM Magazine, Vol. 15 No. 6,
pp. 364-373.
Jiang, G., Ma, F., Shang, J. and Chau, P.Y. (2014), “Evolution of knowledge sharing behavior in social
commerce: an agent-based computational approach”, Information Sciences, Vol. 278, pp. 250-266.
Laureani, A. and Antony, J. (2019), “Leadership and lean six sigma: a systematic literature review”,
Total Quality Management and Business Excellence, Vol. 30 Nos 1/2, pp. 53-81.
Law, A.M., Kelton, W.D. and Kelton, W.D. (2007), Simulation Modeling and Analysis, McGraw-Hill,
New York, NY, Vol. 3.
Lin, C., Baines, T.S., O’kane, J. and Link, D. (1998), “A generic methodology that aids the application of
system dynamics to manufacturing system modelling”, in Proceedings of the International
Conference on Simulation, pp. 344-349.
Maidstone, R. (2012), “Discrete event simulation, system dynamics and agent-based simulation:
discussion and comparison”, System, Vol. 1 No. 6, pp. 1-6.
Michael, G., Rowlands, D. and Kastle, B. (2004), What Is Lean Six Sigma?, McGraw-Hill, New York, NY.
Montevechi, J.A.B., Pereira, T.F., de Carvalho Paes, V., Banerjee, A. and Thomassie, R. (2016), “A study
on the management of a discrete event simulation project in a manufacturing company with
PMBOK®”, Paper presented at the Winter Simulation Conference (WSC).
Paolucci, M. and Sacile, R. (2016), Agent-Based Manufacturing and Control Systems: New Agile
Manufacturing Solutions for Achieving Peak Performance, CRC Press, Boca Raton.
Parunak, H.V.D., Savit, R. and Riolo, R.L. (1998), “Agent-based modeling vs equation-based modeling: a
case study and users’ guide”, jn International Workshop on Multi-Agent Systems and Agent-
Based Simulation, Springer, Berlin, Heidelberg, pp. 10-25.
Pepper, M.P. and Spedding, T.A. (2010), “The evolution of lean six sigma”, International Journal of
Quality and Reliability Management, Vol. 27 No. 2, pp. 138-155.
Pyzdek, T. and Keller, P.A. (2014), The Six Sigma Handbook, McGraw-Hill Education, New York, NY.
Railsback, S.F., Lytinen, S.L. and Jackson, S.K. (2006), “Agent-based simulation platforms: review and
development recommendations”, Simulation, Vol. 82 No. 9, pp. 609-623.
Radzicki, M.J. and Taylor, R.A. (2008), “Origin of system dynamics: Jay W. Forrester and the history of
system dynamics”, US Department of Energy’s introduction to system dynamics (accessed 23
October 2008).
Robert, C.P. (2004), Monte Carlo Methods, John Wiley & Sons, Ltd, Hoboken, NJ. Simulation
Rocha, A.D., Barroca, P., Dal Maso, G. and Oliveira, J.B. (2017), “Environment to simulate distributed methodologies
agent based manufacturing systems service orientation”, in Holonic and Multi-Agent
Manufacturing, Springer, Berlin, pp. 405-416.
Scholl, H.J. (2001), “Agent-based and system dynamics modeling: a call for cross study and joint
research”, in Proceedings of the 34th Annual Hawaii International Conference on System
Sciences, IEEE, p. 8.
Shah, R. and Ward, P.T. (2003), “Lean manufacturing: context, practice bundles, and performance”,
427
Journal of Operations Management, Vol. 21 No. 2, pp. 129-149.
Siebers, P.O., Macal, C.M., Garnett, J., Buxton, D. and Pidd, M. (2010), “Discrete-event simulation is
dead, long live agent-based simulation”, Journal of Simulation, Vol. 4 No. 3, pp. 204-210.
Simon, K. (2007), “SIPOC diagram”, Retrieved January, Vol. 15, p. 2008.
Stamm, M.L., Neitzert, T. and Singh, D.P.K. (2009), “TQM, TPM, TOC, lean and six sigma – evolution of
manufacturing methodologies under the pardigm shift from Taylorism/Fordism to Toyotism?”,
16th International Annual EurOMA Conference, Goteborg.
Sterman, J.D. (2001), “System dynamics modeling: tools for learning in a complex world”, California
Management Review, Vol. 43 No. 4, pp. 8-25.
Stranjak, A., Dutta, P.S., Ebden, M., Rogers, A. and Vytelingum, P. (2008), “A multi-agent simulation
system for prediction and scheduling of aero engine overhaul”, Proceedings of the 7th
International Joint Conference on Autonomous Agents and Multiagent Systems: Industrial
Track, pp. 81-88.
Tesfamariam, D. and Lindberg, B. (2005), “Aggregate analysis of manufacturing systems using system
dynamics and ANP”, Computers and Industrial Engineering, Vol. 49 No. 1, pp. 98-117.
Tolga Taner, M., Sezen, B. and Antony, J. (2007), “An overview of six sigma applications in healthcare
industry”, International Journal of Health Care Quality Assurance, Vol. 20 No. 4, pp. 329-340.
Varga, A. (2001), “Discrete event simulation system”, Proc. of the European Simulation Multiconference
(ESM’2001), pp. 1-7.
Yusof, S.R.M. and Aspinwall, E. (2000), “Total quality management implementation frameworks:
comparison and review”, Total Quality Management, Vol. 11 No. 3, pp. 281-294.
Yusuf, Y.Y., Sarhadi, M. and Gunasekaran, A. (1999), “Agile manufacturing: the drivers, concepts and
attributes”, International Journal of Production Economics, Vol. 62 Nos 1/2, pp. 33-43.
Zagonel, A.A., Rohrbaugh, J., Richardson, G.P. and Andersen, D.F. (2004), “Using simulation models to
address what if questions about welfare reform”, Journal of Policy Analysis and Management,
Vol. 23 No. 4, pp. 890-901.
Zapciu, M. and Cavalieri, S. (2017), “Production systems flow modelling using decomposition method
and required buffers”, International Journal of Simulation Modelling (IJSIMM), Vol. 16 No. 2.

Further reading
Snee, R.D. (2010), “Lean six sigma – getting better all the time”, International Journal of Lean Six Sigma,
Vol. 1 No. 1, pp. 9-29.

Corresponding author
Ali Ahmed can be contacted at: z5007111@unsw.edu.au

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: permissions@emeraldinsight.com

View publication stats

You might also like