You are on page 1of 12

A Review of the Contribution of Hybrid Modelling to

Supply Chain Simulation

R. Y.A. Haddad1, E.C. Heeringa2, S. Jongerden1, P. Schwarz1


1 Faculty of Technology, Policy and Management, Delft University of Technology, Netherlands
{rabeeyacoubayedhaddad,s.jongerden,p.schwarz}@student.tudelft.nl
2 Faculty of Civil Engineering, Delft University of Technology, Netherlands
e.c.heeringa@student.tudelft.nl

Abstract.
In the past decades the supply chain management field has seen an increase of
hybrid modeling approaches that attempt to simulate both macro and micro ef-
fects within the supply chain. Due to the complexity of, and limited experience
with hybrid models, the supply chain field is unaware what the additional insights
of such an approach could be. Therefore, a literature desk research was per-
formed. First a historical overview of hybrid modeling was presented to create
an understanding of the current state of the art. Second, the reason to model sup-
ply chains by means of a hybrid approach was described. Third, the added value
of hybrid modeling within supply chains was introduced, and last, the general
problems with hybrid modeling on a broader scale was described. Results from
this literature review indicate that hybrid modeling, definition wise, has experi-
enced a lot of inconsistencies as multiple simulation and modeling methodologies
were presented as hybrid models. Second, the application of hybrid modeling that
were successful within supply chain management were able to increase model
realism minimalistically, in exchange for a significant increase in model com-
plexity. In a world where models are based on innumerable simplification simu-
lators should question if such an approach could be justified.

Keywords: Supply Chain Management · Simulation · Hybrid modeling · Dis-


crete modeling · Continuous modeling

1 Introduction

Since the need was identified to approach “set[s] of networked organizations work-
ing together to source, produce, and distribute products and services to consumers”
(Lee, Cho, Kim & Kim, 2002, p. 375) as systems, instead of their constituent parts
separately, and the term ‘supply chain’ gained currency, much progress has been made
in modelling such systems. Both continuous and discrete modelling have been used
extensively. However, not all aspects of supply chain management can be appropriately
represented by either discrete or continuous models. To investigate such problems, hy-
brids of discrete and continuous with various degrees of integration have been made.

adfa, p. 1, 2011.
© Springer-Verlag Berlin Heidelberg 2011
The developments in hybrid modelling over the last few years and the effort made
to construct a full hybrid modeling approach have raised questions on its feasibility.
The supply chain community and other fields are left with the question of how to ap-
proach ‘hybrid’ problems. To assist the field of SCM, this paper will try to shed light
on the need for and practicalities and shortcomings of a full hybrid modeling approach
within supply chain analysis by means of the following research question:

“For which aspects in supply chain management can hybrid modelling provide ad-
ditional insights compared to traditional decoupled discrete and continuous model-
ling approaches?“

To answer this research question we will first define supply chain management and
the use of simulation in the study of supply chains. Second, we will briefly present the
essence of discrete and continuous modeling techniques and indicate key differences
between the two approaches. Third, we will go more deeply into the motivation for
hybrid models, providing an expedient definition. In addition we will outline techniques
to implement such hybrid models in SCM simulation. Fourth, we will present chal-
lenges from a modelling perspective to create fully integrated hybrid models. Fifth, we
will reflect on the use of hybrid modeling techniques within Supply chain management.
Finally, we will conclude with a discussion of the contribution of hybrid modeling to
SCM and indicate opportunities for further research.

The research methodology selected for this research, in which mostly emphasis was
placed on the theoretical understanding of hybrid modeling within supply chain analy-
sis, is desk research of scientific articles. Scientific literature from both the simulation
and supply chain field were studied to create an understanding if hybrid models would
be required within supply chain analysis. Furthermore, studying these two fields might
provide room for suggestion for further research.

2 A Historical Overview of Hybrid Supply Chain Modeling

The first efforts to optimize the design and operation of supply chains were based on
the application of linear and nonlinear optimization techniques. Typically an objective
function is optimized for variables of interest such as cost or profits, given certain con-
straints. Solving optimization problems by means of linear and nonlinear solvers im-
plied that problems could only be of limited complexity and had to be static in nature.
Dynamic programming provided a solution for the mismatch of the dynamic nature of
supply chains and their static models by allowing stochastic demand to be taken into
consideration. However, none of the optimization techniques were able to deal with
issues such as forecasting errors. This and other typical problems in SCM, such as the
complexity of models, mathematical queuing theory limitations and cost related limi-
tations (Thierry, Bel & Thomas, 2010) resulted in a shift towards simulation techniques.
Simulation techniques created new opportunities to assess forecasting errors and to an-
alyze more complex supply chains as well as the effectiveness of supply chain policies
(Ingalls, 1998).
For the analysis of forecasting errors on the supply chain and especially effects from
tactic and strategic changes, in most cases discrete-event simulation were used. Follow-
ing the taxonomy of Nance (1981) discrete-event simulation concerns “the representa-
tion of a system as it evolves in time in which state variables change instantaneously at
[events in time]” (Lee, Cho, Kim & Kim, 2002, p. 379). As a result of the use of dis-
crete-event simulation in supply chain analysis, there was no reflection on the continu-
ous nature of the process, as various aspects of supply chains are neither completely
discrete nor completely continuous. Therefore, shifts to hybrid models within supply
chain analysis were endeavored as such models might provide more information re-
garding the micro and macro level of the system, wherein hybrid is referred to as a
combined discrete/continuous modeling approach (Lee, Cho, Kim & Kim, 2002).

This hybrid modeling approach was already introduced many years before the SCM
community became interested in its application. Fahrland (1970) questioned in 1968
why one should limit research to either a discrete or continuous modeling approach,
especially when certain situations required a more interdisciplinary method. Fahrland’s
hybrid modeling approach was a composition of continuous and discrete sub-systems,
which were simulated concurrently by means of communicative interaction. By devel-
oping this method, Fahrland resolved only one part of the problem as the remaining
question regarding the simulation language was left open (Fahrland, 1970).
The simulation languages that were developed for discrete modeling after Fahrland
introduced his new approach were primarily focused on implementing primitive con-
tinuous simulation capabilities into discrete event languages. Some of these languages
were: GASP IV, DISCO, SLAM II, COSMOS and CSSL. Barton and Pantelides (1994)
argued however that these simulation languages fall short when applied to complex
processing systems. Furthermore, they commented on Fahrland’s approach to integrate
discrete and continuous simulation languages. First, they substantiate this approach and
describe it as “(…) particularly suitable for those systems in which multiple entities
with identical continuous behaviors periodically enter and leave the overall system.”
(Barton and Pantelides, 1994, p. 970). However, later in the same paper they argue that
processing systems presenting discontinuity in the fundamental physical behavior or
physicochemical discontinuities could neither be viewed as purely discrete nor as
purely continuous. The decomposition of a model in a discrete subsystem, representing
external actions, and a continuous subsystem, partially representing physical behavior,
seemed rather illogical. Even so, they did not provide a solution to this problem (Barton
and Pantelides, 1994).
When Brailsford, Desai and Viana (2010) reviewed the state of the art in 2010 they
mentioned that not much progress had been made over the years to develop a genuinely
hybrid modeling system that combines both discrete and continuous languages. They
stressed that it was not the problem to develop the software but rather the absence of a
conceptual philosophy and a practical methodology.
3 Supply Chain Management

CSCMP defines SCM as “… an integrating function with primary responsibility for


linking major business functions and business processes within and across companies
into a cohesive and high-performing business model” (CSCMP, Stadtler & Kilger, p.
11). Others put less emphasis on the links and more on the flows of material, infor-
mation and money, e.g. Stadtler & Kilger (2010, p.11): ‘[SCM is] the task of integrating
organizational units along a supply chain and coordinating material, information and
financial flows in order to fulfil (ultimate) customer demands with the aim of improving
the competitiveness of a supply chain as a whole.’
Generally SCM is understood to concern the ‘organizational aspects of integrating
legally separated firms as well as coordinating materials, information and financial
flows within a production-distribution network’ (Stadtler & Kilger 2010). As firms
have followed a trend towards specialization in the past decades, the legal separation
between actors within supply chains has increased. Whereas large corporations con-
trolled much of their supply chains in the past, modern supply chains have become far
more diversified from an organizational perspective. Furthermore, products and ser-
vices have become more complex, which has further increased the number of actors in
the supply chain. The trends of specialization and increasing structural complexity have
made SCM more relevant than ever.
Modelling and simulation are becoming increasingly important to tackle complex
problems in Supply Chain Management (SCM) that cannot be solved with linear opti-
mization. Therefore, computational models are used as means for decision support tools
across many aspects of SCM. Findings of modelling & simulation are manifold: (1)
They may provide an enhanced understanding of the internal mechanism of the supply
chain, (2) test responses to intervention and (3) allow to compare system performance
in a wide range of scenarios (Sweetser, 1999). Typically, “what if” analysis are per-
formed to identify the optimal or most robust alternatives from a set of policy options
(Tako & Robinson, 2012).

More to add

4 Discrete and Continuous Modeling

A variety of modelling formalisms exist, which each have distinct characteristics and
typical application areas. Fundamentally, there is a division between discrete event
modelling and continuous system dynamics (SD) modelling.
System dynamics models describe systems by depicting causal relations between
factors. The building blocks are stocks, flows and arbitrary variables. State change oc-
curs continuously over time, however, in order to solve the differential equation (ap-
proximately, by means of a numerical method) time is discretized. Generally SD mod-
els are deterministic and represent averages. According to Jay W. Forrester, the founder
of SD, central concepts of SD approach are the different types of delay function and the
disparate behavior of positive/negative feedback loops (Forrester, 1961). In contrast,
discrete event models describe systems as networks of queuing system and activities.
The state is constant between events, which occur at irregular instants of time. In con-
trast to SD models, discrete event models are stochastic in nature, generated through
the use of statistical distributions and random variables.
Generally it is claimed that SD modelling is suitable for addressing problems at a
strategic level, whereas discrete event modelling is more applicable for problems at an
operational or tactical level. A strategic decision making process is typically character-
ized by minimum data availability, deep uncertainty, qualitative parameters and behav-
ior at of a continuous nature at an aggregated level (Helal et al., 2007). SD is suitable
for clarifying patterns of behavior and system structures and can capture the long term
effects of decisions. Further, it helps to shed light on the interaction and interdepend-
encies between decisions made by different actors within a system, (e.g. different firms
across a supply chain) and softer (less commonly quantified) aspects relevant to deci-
sion making (e.g. human behavior). Finally, SD may help to address issues such as
system stability (Rabelo et al., 2005). Hence, the popularity of SD in strategic studies.
In contrast, discrete event modelling is preferably used for problems with a narrow
scope, where specific optimization problems are of interest (Rabelo et al., 2005). There-
fore discrete event modelling is generally used for problems at an operational or tactical
level, where high-level parameters are not part of the decision space. However, a liter-
ature review by Tako and Robinson (2012) suggests that characteristics other than stra-
tegic, tactical or operational of the issue at stake determine the choice between SD and
discrete event modelling in supply chain modelling.
Tako and Robinson conducted a review of 127 papers, out which 86 (68%) used a
discrete event approach, 38 (30%) an SD approach, while only three (2%) used a com-
bination of discrete event and SD modelling. They found that the topics of the papers
were difficult to subdivide into strategic, tactical and operational issues because often
several levels of decision-making are involved. Rather than doing so, Tako and Robin-
son made their own subdivision into fourteen issues on the spectrum between the ex-
tremes of strategic issues on the one end and operational issues on the other end. The
following fourteen categories were defined ranging from strategic to operational:

supply chain structure; process redesign; supplier selection; facilities/capacity


planning; replenishment control policies; bullwhip effect; information sharing;
supply chain integration; supply chain optimization; cost reduction; system per-
formance; inventory planning/management; planning and forecasting demand;
production planning and scheduling; distribution and transportation planning;
dispatching rules; reverse logistics.

Papers could be classified under more than one category. Instead of a clear dichot-
omy between discrete event modelling used for operational issues and SD for strategic
issues, Tako and Robinson found that the use of both techniques was fairly evenly dis-
tributed between both extremes. However, discrete event modelling was found to be
particularly popular in research on system performance, inventory planning/manage-
ment and production planning & scheduling. SD modelling is most common in studies
of information sharing, bullwhip effect and inventory planning/management, domains
of supply chain management in which SD has traditionally been influential (e.g. Ster-
man, 1989). The three papers in which hybrid modelling was used were classified under
topics across most of the spectrum from strategic to operational issues, but all three
papers were classified under production planning & scheduling. Therefore, based on
Tako and Robinson’s research, there is no reason to conclude that hybrid modelling is
of use in one particular domain of the vastly diverse field of supply chain management.

5 Hybrid Modeling and the Levels of Integration

Hybrid models have attracted more and more attention in recent years for several
reasons, of which we highlight three. First, the current state of SCM is characterized by
increasing levels of integration across different enterprises across the supply chain. Sec-
ond, other aspects than the material flow and their interaction are recognized of growing
importance. Third, supply chains are increasingly seen in their context as a changing
business environment, driven by dynamics regarding competition, changing customer
preferences and the general economic situation. Fahrland stated that when the main
interest is the interaction between discrete and continuous elements, hybrid models are
required (Fahrland, 1970). Such interaction calls for a comprehensive hybrid model that
allow to encompass differences in decision making levels, scope, and frequency of de-
cision making, required detail and the accuracy needed at the moment of decision mak-
ing (Helal et al., 2007).
Based on this growing attention to hybrid modelling, it is necessary to consider the
definition of hybrid modelling. What makes this even more important is the fact that
researchers have used different definitions for hybrid modelling simultaneously, as they
refer to hybrids when discussing combined, coupled and distributed simulation tech-
niques. Generally, definitions agree on the basic premise that hybrid models include the
strategic character inherent to continuous modeling and the operational processes that
are characteristics of discrete modeling, however, they use various definitions that are
describing the different levels of integration of discrete and continuous modelling rather
than a consistent definition of hybrid modelling. This has raised a flag to the necessity
of making a clear distinction between the definition of hybrid modelling and the levels
of integration of such approach.
A first definition considers hybrid modelling as a method where two components are
distinct but and simply exchange information. Cahal and Eldabi (2008) state that hybrid
models function like bouncing models in which information is bounced between the
continuous and discrete components of the models.
A second definition from Barton and Lee (2002) describes hybrid models as “a [col-
lection of] discrete state subsystem[s], a [collection of] continuous state subsystem[s],
and the possible interactions between these subsystems” (Barton and Lee, 2002, p. 258).
This implies that there are two different models that communicate with each other. This
definition can be related to the interchangeable and misleading use of the term ’hybrid
modelling’ and distributed simulation. The latter is characterized by loosely coupled
simulations that interact intensively at certain points in time (Helal et. al, 2007). The
two are certainly related in their approach of interaction between two separate parts.
However distributed is more related to simulation than to a modelling technique, I.e.
modelers create their separate models using any suitable modelling technique, and
when simulated, the interaction occurs through different architectures like High Level
Architecture (HLA).
A third definition defines a hybrid model as an integrated model that incorporates
both discrete and continuous variables within the same model. (Lee et al., 2002). This
definition can be related to Ziegler’s work in combining DEV and DESS into DEVS.
Using this approach, the discrete and continuous components consisting of sets of in-
puts, outputs, states, and the transition, output, and rate of change functions in their
original formalisms are transformed into a unified format to specify the hybrid system.
(Ziegler, 2006). Defining a hybrid model following Ziegler’s approach implies that the
components are more deeply integrated as both are transformed into a common ap-
proach, enabling communication to happen as if the two are actually one model.
This diversity of definitions creates a necessity to differentiate between the defini-
tion of hybrid models and the levels of integration used to combine discrete and con-
tinuous models. Chahal and Eldabi (2008) have identified three modes for such inte-
gration, which can be a starting point for a more elaborative guideline for making this
distinction. The simplest of the three is the hierarchical mode, where both models are
marked and simply exchange data between each other which can be related to the first
group of definitions. The second mode is the process – environment, where both sys-
tems are distinct, but the discrete model, representing processes, is part of the continu-
ous model representing the environment. Through this mode there is a deeper level of
interaction that goes beyond a simple data exchange between the two, which is closely
related to the second group of definitions. The final mode is the integrated mode, where
there is no distinction between the discrete parts and the continuous parts. Ziegler ap-
proach and the definitions related to it can be related to this mode, even though some
might argue that it is not fully integrated but it is a much deeper integration than the
previous two.
Finally, it is worth noting that the deeper the level of integration, the more complex
modeling can get. Consequently, it is essential to justify the depth of integration re-
quired with the objectives of the simulation study before diving into the complexities
of full integration.

6 The Current Difficulties to Construct a Full Hybrid


Methodology

Since the introduction of the idea of integrated continuous and discrete subsystems
by Fahrland in 1970, the simulation field has adopted this approach as a ‘combined
discrete/continuous approach’ or ‘hybrid’ approach. Since the beginning scholars have
criticized this approach on various aspects (Barton and Pantelides, 1994; Vangheluwe,
De Lara and Mosterman, 2002; Johansson, Lygeros and Sastry, 2004; Brailsford, Desai
and Viana, 2010). Some of the most heavily discussed problems will be discussed in
the following section.
Barton and Pantelides (1994) argued that the composed continuous and discrete
subsystems methodology from Fahrland (1970) was illogical as it would be unable to
model physicochemical discontinuities. These physicochemical discontinuities occur
in fundamental physical behavior and are mostly found in thermodynamics, fluid me-
chanical transitions, and irregularities in geometry and are considered an essential ele-
ment of physical behavior. An example of a physicochemical discontinuities is a flow
transition in fluid mechanics where one could not determine a priori which set of equa-
tions to use due to the absence of information. This is further explained by Rico-
Ramirez, Allan and Westerberg (1999, p. 520) as:

“…the representation of conditional models is particularly difficult because


it is necessary to declare alternative modeling equations which become active
depending on the values of the modeling variables. Moreover, the solution to
conditional models involves simultaneously selecting the equations to be
solved and solving them.”

In a composed methodology this would pose a problem as it is unknown a priori


what type of calculation must be performed and therefore could not be pre-assigned to
a discrete or continuous sub-model (Barton and Pantelides, 1994).
A second problem of integrating continuous and discrete subsystems was in-
dicated by Vangheluwe, De Lara and Mosterman (2002) and later was substantiated by
Brailsford, Desai and Viana (2010), who argued that sub-models may be inconsistent,
since information from lower levels of abstraction are not available on the aggregated
problem level. Objective inconsistencies occur when, due to aggregation, one is unable
to decompose higher levels of abstraction into lower levels of abstraction. Brailsford,
Desai and Viana (2010) give an example of merging individual entities into a continu-
ous mass and then wanting to split this continuous mass back into the original popula-
tion, based on what happened to the mass during the continuous time period. The prob-
lem of objective inconsistency would be even more difficult to resolve if one would
take into account the possible omission of entities from the lower abstraction level,
which therefore cannot be represented in the higher level of abstraction (Benjamin, Er-
raguntla, Delen & Mayer, 1998).
Besides the aforementioned problems of Fahrland’s methodology, problems of a
mathematical nature were identified. Johansson, Lygeros and Sastry (2004) describe
some of the mathematical problems that occur when hybrid modeling is performed in
a computer simulation model. The first problem they mention is the lack of continuity
of the solution with respect to the initial conditions. Second, due to the nature of hybrid
computation, hybrid systems can potentially perform an infinite number of discrete
transitions in finite time. This problem is also referred to as the Zeno phenomenon.
These and other mathematical problems could be solved only recently by the develop-
ment of the DEV and DESS in DEVS framework by Zeigler (2006) where he intro-
duced the ‘state event detection’ and the ‘integrator state reset’.
Nevertheless, the time scales inconsistency between discrete and continuous models
remains. Even though the problem of combining two time scales has been solved math-
ematically (Belbas, 2004), this does not mean that is does not pose a problem for actu-
ally running simulations with such a model. There is simply a fundamental time differ-
ence between discrete and continuous components, in which for instance the discrete
sub models have a timescale of hours, days or weeks, whereas the continuous sub mod-
els have a timescale of months, years or even decades. Of course there is a simple so-
lution to this problem, which is running discrete models for a very long time, but this
would not be desirable from a computational perspective.
Even though the compromises and the remaining problems currently prevent the
creation of full hybrid models, research is advancing towards the validity of full hybrid
modelling.

7 Are the Current Techniques Sufficient to Model Supply


Chains Efficiently?

Throughout the different parts of the supply chain, processes differ with respect to
their continuous or discrete characteristics. Due to the diverse nature of supply chain
processes, supply chains can be divided into different areas according to the nature of
operations, variables and dynamic nature of each area. Ingalls (1998) splits the supply
chain into three levels; scheduling, tactical planning and strategic planning. This dis-
tinction of modeling approaches for supply chain aspects was also found in a literature
study by Tako and Robinson (2011). They performed an analysis on the application of
discrete, continuous and hybrid modeling techniques to simulate aspects of supply
chains, ranging from a strategic level, through a tactical level to an operational level.
Tako and Robinson expected that studies in which SD modelling was used would focus
primarily on the strategic level, whereas studies that used discrete modelling would
focus more on the operational level. Their hypothesis was proven to be significant and
they concluded that “SD has a slightly more strategic focus” (p.807), although the sig-
nificance of the relation between SD modelling and strategic issues was due to the large
number of SD simulations used to simulate the ‘bullwhip effect’. When the studies
concerning the bullwhip effect were removed from the data, no significant dichotomy
between the use of discrete and continuous modelling with respect to operational versus
strategic issues could be found.
Assuming that this relation holds, one would expect that hybrid models would posi-
tion themselves between a strategic and tactical modeling approach, as they combine
these distinct parts of the SC in one model (Lee, Cho, Kim & Kim, 2009; Lee, Cho &
Kim, 2002; Venkateswaran & Son, 2012; Tako & Robinson, 2012). Although, research
from Tako and Robinson (2011) indicated (although very little evidence) that hybrid
models are used over the entire range from tactical to strategic. Tako and Robinson
(2011) did however refer to hybrids as a combined modeling approach while the authors
they depicted on using hybrid models, referred to hybrids differently. Under the hybrid
modeling approach they either meant coupled or distributed simulation.
The papers referred to by Tako & Robinson (2011; Venkateswaran and Son, 2005;
Lee, Cho and Kim, 2002; and Reiner, 2005) all indicated to have some success with the
‘hybrid’ approach to simulating supply chains. Although that benchmarks in compari-
son to either discrete or continuous modeling approaches still had to be performed. In
overall, one could say that approaches in which multiple modeling languages are com-
bined lead to more insights into system performance, as was also mentioned by Barton
and Pantelides (1994). Indicated by literature (although little evidence) the use of hy-
brid modeling appear not to depend on if one wants to perform a strategic or tactical
level analysis. Researchers tend to use the modeling language they are most accustomed
to and add the discrete or continuous part.
Even so, the use of these hybrid modeling techniques seem to be adding value and
create more understanding than single modeling approaches, even when these ap-
proaches are not considered to be fully hybrid (Barton and Pantelides, 1994). But since
supply chains do not portray physicochemical discontinuities as these are only present
in thermodynamics, fluid mechanical transitions, and irregularities in geometry, one
could question the need for a full hybrid modeling approach for this sector. From papers
such as Venkateswaran and Son (2005), Lee, Cho and Kim (2002) and Reiner (2005)
it becomes clear that the approaches available are proving to be sufficient to simulate
supply chains effectively.

8 Conclusion

Within supply chain simulations, researchers have tried to represent both the tactical
discrete processes and strategic continuous processes in one model. Due to the limita-
tions of the modeling methodologies available, hybrid models were used. Hybrid mod-
els have proven to be a modeling methodology that can add value to discrete and con-
tinuous simulation studies, without there currently being a particular aspect within SCM
that especially requires the use of hybrid models and they can therefore, based on the
current experience, add value to all section of SCM. Even though these successful im-
plementations, one could question if the increase in complexity, resulting in only a very
small increase of model realism, can be substantiated in a world where models are based
on innumerable simplifications. This poses the question whether the selection of hybrid
modelling can be justified if the same problem may be modelled by a single modelling
technique.

Within the simulation field there is a common misconception on hybrid models as


different definitions are used. Researchers use the terms hybrid for coupled, distributed
and combined modeling approaches, which distinctively have different implications.
Until now, no full hybrid modeling approach has been created due to the absence of a
conceptual philosophy and modeling methodology. But one could question if such an
approach is required in the simulation field as different modeling techniques (distrib-
uted, coupled and DEVS) are proving to be very effective approaches of combining
discrete and continuous languages.
9 Further Room for Research

To prevent further improper use of the term ‘hybrid’, further research should be per-
formed to construct a clear and accepted definition. This would then ensure a better
understanding of what hybrid models are and prevent future misuse and confusion when
multiple authors refer to hybridity. This thought follows Nance’s (1981) as the estab-
lishment of fundamental definitions for the simulation domain realigns researchers and
improves discussion on simulation subjects.
Furthermore, due to the low number of practical applications of hybrid models in
SCM or in other sectors compared to the application of traditional methods, there is
confusion on when to use hybrid models. There is a desire for a deeper understanding
on the optimal balance between the added value of hybrid models and their effect on
the model complexity. Therefore, further research should be performed to establish a
threshold that assists researchers in selecting traditional modeling techniques or hybrid
modeling techniques.

10 References
1. Bagchi,S., Buckley, S., Ettl,M., Lin,G., (December 1998) Experience using the IBM supply
chain simulator, in: Winter Simulation Conference
2. Barton, P. I., & Pantelides, C. C. (1994). Modeling of combined discrete/continuous pro-
cesses. AIChE journal, 40(6), 966-979.
3. Barton, Paul I., and Cha Kun Lee. 2002. Modeling, simulation, sensitivity analysis and op-
timization of hybrid systems. ACM Trans-actions on Modeling and Computer Simulation
4. Beamon BM. (1998) Supply chain design and analysis: models and methods. International
Journal of Production Economics ;55(3):281–94
5. Belbas, S. A. "Dynamic programming in mixed continuous-discrete time scales." arXiv pre-
print math/0408378 (2004).
6. Benjamin, P., Erraguntla, M., Delen, D., & Mayer, R. (1998, December). Simulation mod-
eling at multiple levels of abstraction. In Proceedings of the 30th conference on Winter sim-
ulation (pp. 391-398). IEEE Computer Society Press.
7. Brailsford, S. C., Desai, S. M., & Viana, J. (2010, December). Towards the Holy Grail:
combining system dynamics and discrete-event simulation in healthcare. In Simulation Con-
ference (WSC), Proceedings of the 2010 Winter (pp. 2293-2303). IEEE.
8. Chahal, K. & Eldabi, T., (2008). Applicability of hybrid simulation to different modes of
governance in UK healthcare. Proceedings of the 2008 Winter Simulation Conference S. J.
Mason, R. R. Hill, L. Mönch, O. Rose, T. Jefferson, J. W. Fowler eds.
9. Fahrland, D. A. (1970). Combined discrete event continuous systems simulation. Simula-
tion, 14(2), 61-72.
10. Helal, M., Rabelo, L., Sepulveda, J. and Jones, A. (2007), “A methodology for integrating
and synchronizing the system dynamics and discrete event simulation paradigms”, Proceed-
ings of the 25th International Conference of the System Dynamics Society, Boston, MA,
Vol. 3 No. 3, pp. 1-24
11. Ingalls, R. G. (1998, December). The value of simulation in modeling supply chains. In
Proceedings of the 30th conference on Winter simulation (pp. 1371-1376). IEEE Computer
Society Press.
12. Johansson, K. H., Lygeros, J., & Sastry, S. (2004). Modeling of hybrid systems.
13. Lee, Y. H., Cho, M. K., & Kim, Y. B. (2002). A discrete-continuous combined modeling
approach for supply chain simulation. Simulation, 78(5), 321-329.
14. Lee, Y. H., Cho, M. K., Kim, S. J., & Kim, Y. B. (2002). Supply chain simulation with
discrete–continuous combined modeling. Computers & Industrial Engineering, 43(1), 375-
392.
15. Nance, R. E. (1981). The time and state relationships in simulation modeling. Communica-
tions of the ACM, 24(4), 173-179.
16. Reiner, G. (2005). Customer-oriented improvement and evaluation of supply chain pro-
cesses supported by simulation models. International journal of production economics,
96(3), 381-395.
17. Rico-Ramirez, V., Allan, B. A., & Westerberg, A. W. (1999). Conditional modeling. 1. Re-
quirements for an equation-based environment. Industrial & engineering chemistry research,
38(2), 519-530.
18. Tako, A. A., & Robinson, S. (2012). The application of discrete event simulation and system
dynamics in the logistics and supply chain context. Decision support systems, 52(4), 802-
815.
19. Thierry, C., Bel, G., & Thomas, A. (2010). The role of modeling and simulation in supply
chain management. SCS M&S Magazine, 1(4), 1-8.
20. Vangheluwe, H., De Lara, J., & Mosterman, P. J. (2002, April). An introduction to multi-
paradigm modelling and simulation. In Proceedings of the AIS’2002 conference (AI, Simu-
lation and Planning in High Autonomy Systems), Lisboa, Portugal (pp. 9-20).
21. Venkateswaran, J., & Son*, Y. J. (2005). Hybrid system dynamic—discrete event simula-
tion-based architecture for hierarchical production planning. International Journal of Pro-
duction Research, 43(20), 4397-4429.
22. Zeigler, B. P. (2006, April). Embedding DEV&DESS in DEVS. In Proc. DEVS Integrative
M&S Symp.

Not complete yet

You might also like