Modularity in System Dynamics: representing
closed-loop supply chain configurations
Aly Elmasry Andreas GroBler
European Masters programme in Department of Operations Management,
System Dynamics University of Stuttgart
alyelmasryO@gmail.com andreas.groessler@bwi.uni-stuttgart.de
Abstract
Modularity has been used in several disciplines as a structured way to approach
complexity, to increase efficiency, to reduce repetitive work, to decrease errors, and to
create cumulative learning. The system dynamics field has yet to benefit from
modularity as a systematic method for developing models. In this article, we discuss
system dynamics modules that can be used in developing supply chain models. First,
certain design rules are identified for developing modules in system dynamics. Then, a
theoretical framework is used to guide closed-loop supply chain module development.
Afterwards, reoccurring structures are identified to be used in developing system
dynamics modules for modelling supply chain systems. Finally, the modules were used
to represent three different closed-loop supply chain configurations; we show the
implications of each configuration.
Introduction
Developing system dynamics models is an extensive process that involves identifying
problem variables, developing system structure, and analysing dynamic behaviour,
along with validation tasks on each of these steps (Sterman, 2000: Chapter 3).
Modularity simplifies this process, by providing ready-made modules that could be
plugged together to represent the system under study. Although there have been
previous attempts to develop generic model building blocks for model development, for
example the system dynamics molecules (Hines, 1996), there are no attempts to apply
modularity systematically. In addition, no application specific modules were developed
in the system dynamics field, for instance for supply chain applications.
Most of system dynamics models in supply chain are developed using similar
structures, for example: material flows, stock control heuristics, incoming orders
processing, costs and revenue, and information flows, all of which could be represented
1
Modularity in System Dynamics:
i d-loop supply chain confi
by generic modules. This motivated us to apply modularity to supply chain
applications; as Wolstenholme and Coyle (1983: 2) put it:
“The modular approach arises from the recognition that systems
in a wide variety of fields contain structurally similar
elements...”
To apply modularity in supply chain modelling, we use Wikner and Tang's (2008)
theoretical framework for closed-loop supply chain configurations. In addition, we
review system dynamics models in supply chain management, to deduce reoccurring
structures and use them in developing system dynamics modules for supply chain
applications. Upon developing supply chain modules in system dynamics, we perform
several validation tests to eliminate the possibility of modelling errors and increase
confidence in the modules’ structures and their behaviour (Forrester & Senge, 1980).
Lastly, we use these modules to represent three different closed loop supply chain
configurations and show the implications of each configuration.
The benefits of modularity
Modularization as a practical concept can be credited to the computer industry (Baldwin
& Clark, 1997) as a way to break up computer programs into separate components. The
components (i.e. modules) follow certain design rules and standards that allow them to
be coupled via predefined interfaces. Baldwin and Clark (2000) argue that the power of
modularity lies in the structurally successful partition of a system into blocks or
modules that can be developed independently and in parallel, and then integrated
together to form the whole.
Nowadays, the high promise of modularity is recognized across many different fields.
Modular-construction is used today to apply standardization and predesign leading to
mass production, significant reduction in on-site construction, process efficiency, and
the ability to repeat orders for clients (Gibb, 2001). In the computer industry,
modularization allowed for independent experimentation and design of components,
leading to innovation and high rate of advancement (Baldwin & Clark, 1997). In
software development, modularization allowed for ease of work distribution among
software developers; fewer errors while developing software; ease of error recognition;
work structuring; ease of code reading; and the ability to work in parallel (Mall, 2009).
Modularity in System Dynamics:
it d-loop supply chain
Modularity can benefit system dynamics modelling practice in many ways. Model
construction is easier and more efficient when done using previously developed
modules. The reuse of well-validated and tested modules decreases modelling time,
errors and costs. The modelling time decreases by dealing with complexity structurally
through a ‘divide and conquer’ strategy. This allows the modeller to think
systematically and to structure the modelling process. Testing modules individually
could then identify errors in models. When errors are identified, and fixed, the modules
could be used repeatedly in different models: helping the modeller in constructing
future models faster. Furthermore, the modelling software could support class libraries,
so that a structural change in a module class automatically changes all the
corresponding modules in the constructed model. This will further decrease time and
effort in model development and simplify adaptation.
Modules also help throughout the model validation process, in that each module would
have been tested and used several times in distinct models to represent a variety of
systems. When modules are used to represent multiple systems they would pass the
“‘family-member’ test, which is used to “show that the model takes on the characteristics
of different members of the class” where a class is a family of systems (Forrester &
Senge, 1980: 25).
Modularity enhances cumulative learning by improving and building on modules. As
such, modellers can share their developed modules, and others could use and improve
on them. Moreover, modularity allows different modellers to work in parallel when
representing one complex system. Furthermore, modellers develop an_ intuitive
understanding of the modules’ functions and structures after using them repeatedly.
This allows the modellers to see a complex system represented in modules and identify
each module’s function while understanding the overall dynamics of the model. Thus,
modularity simplifies complex models into multiple distinct modules, and makes it
easier for the modeller to represent the system, and understand it.
Obviously, these benefits come with some limitations. First, there is the potential
danger that the ease of modelling that comes with modularity can create false
confidence in some of the models. Even though modules should be validated and tested
thoroughly, the overall model has to be validated again structurally and behaviourally.
Second, the use of modules may prevent the modeller from seeing the endogenous
3
Modularity in System Dynamics:
i d-loop supply chain confi
dynamics of the bigger system. If the model is represented in modules, the main loops
in the overall system could be hidden between the modules. This can be misleading as
endogenous dynamics are one of the most important foundations in system dynamics
modelling (Richardson, 2011). However, modules interact with each other and if one
follows the rigorous system dynamics validation practices (Forrester & Senge, 1980),
these endogenous dynamics should be present in the bigger system. Although
representing a complex system in modules is easier to understand, the modeller
nevertheless needs to present additional representations of the system that show the
main dynamics (e.g. causal loops or stock and flow diagram representing the main
loops). Finally, modules may prevent the deeper understanding that comes with one’s
model development journey. Thus, one should study the modules thoroughly before
using them.
Supply chain modules
System dynamics has a successful history of modelling forward supply chains starting
from Forrester's (1961) seminal book “/ndustrial dynamics”. Angerhofer and Angelides
(2000) categorized system dynamics work in supply chains into three main lines of
research: theory building, problem solving and improving the system dynamics
approach. Interestingly, most of the models discussed use similar components to model
supply chain systems, for example: ordering policy, capacity expansion, production
rate, backlogging orders, product shipment and acquiring rate. Obviously, the models
are structurally different to represent the system at hand. A lot of work is put to develop
these models, and most of them are developed from scratch. Repeated structures are,
often, presented by citing previous work. For example, Georgiadis and Besiou (2008:
1669) cite Sterman (1989) for their production ordering rate structure; and Kamath and
Roy (2007) cite Forrester (1968) for similarities in model structure. Even if structures
within a model are not explicitly cited from previous work, they often retain significant
similarities with previously developed models. This shows that supply chain modules
can be advantageous; when repeating structures are put in modules, modellers could use
the modules frequently while saving considerable time and effort.
Theoretical framework and modules formulation
Complex systems consist of many elements. Elements’ interconnections and interplay
form the complexity and holism of the system (Baldwin & Clark, 2000). The central
4
Modularity in System Dynamics:
it d-loop supply chain
idea to modularity according to Baldwin and Clark (2000) is to have system elements
depend on pre-defined design rules rather than on their interconnections and interplay.
This allows for independent design of modules. This leads to the following design
guideline:
Design guideline 1: Modules should depend on pre-defined
design rules.
Many design rules could be set depending upon modules' applications, functionality,
and usage. As a result, design rules should be specified to guide module design and to
achieve modular usefulness while retaining the module’s applicability in the
corresponding field and application.
Foster (1995) provides a list of recommendations that deem useful as a framework in
modular design in parallel programming. Since software development is closely related
to modelling and simulation, we have extracted the following design guidelines as a
framework for modular design:
Design guideline 2: Each module should have a defined purpose
with no replication of purposes among modules.
Design guideline 3: Interface between modules should be
clearly defined
Design guideline 4: Modules should be hierarchical, and build
on each other.
Design guidelines 1-4 provide the general framework for designing system dynamics
modules for supply chains. The design guidelines will ensure that each module could be
developed with little knowledge about other modules, and allow for intuitive and easy
assembly of modules to fit the purpose of modelling.
To divide the supply chain into interchangeable modules, one has to closely inspect the
structure of the supply chain in system dynamics models. At first glance, it is easy to
recognise that the supply chain could be divided into distinct facilities and actors:
manufacturing centres, distributers, retailers, customers, acquisition centres, and
recovery centres. However, these facilities may overlap in terms of the process being
performed (i.e. violating design guideline 2). For example, a retailer may essentially
Modularity in System Dynamics:
i d-loop supply chain conf
perform the same process like a distributer. At the same time, different retailers may be
different structurally; for example, with regards to their ordering decision. This is due to
the high variance and complexity in supply chain structures.
In supply chain management, high complexity has been dealt with through buffers
(Wikner & Tang, 2008). These buffers can be inventory or time. The buffers allow the
supply chain to be decoupled, as long as these buffers are adequate and meet certain
criteria. The different processes, before and after the buffer, can be designed separately.
This gave rise to the Customer Ordering Decoupling Point (CODP) concept. CODP,
also known as penetration point, is defined as
“the point in the value chain for a product where the product is
linked to a specific customer order.” (Olhager, 2012: 38)
The positioning of CODP divides the supply chain into make-to-stock, make-to-order,
assemble-to-order and engineer-to-order configurations. In make-to-stock
configurations the stock of inventory creates a buffer between customer orders and the
manufacturing process. The manufacturing process before the buffer is forecast driven
rather than demand driven. An example of that is purchasing a product from a retailer
(be it jeans, TV set, or car) and receiving it instantly from stock. This is in contrast to
make-to-order, in which making/manufacturing the product/service is directly
connected to the customer specific order. An example of that is purchasing airplane
engines, custom-build computers, or high-end medical equipment. Assemble-to-order is
essentially assembling the product to customer orders. The extreme opposite of make-
to-stock is engineer-to-order, where the design and engineering of the product is only
initiated with a customer order. For example, custom-build houses, refinery producers,
or wedding-events. Figure 1 shows different possible configurations of the supply chain
according to CODP position. From Figure 1, we can see that CODP divides the supply
chain process to either forecast driven or demand driven processes (Olhager, 2012).
Modularity in System Dynamics:
it d-loop supply chain
Engineer Manufacture Assemble Deliver
Make-to-stock [ Forecast driven }
Assemble-to-order[ Forecast driven e7| Demand driven }
[ Demand driven }
Engineer-to-order [ Demand driven }
Figure 1: Different configurations of supply chains according to CODP position, based on Olhager
(2003)
Make-to-order Forecast driven
Wikner and Tang (2008) provide a framework dependent on CODP for closed-loop
supply chains. The framework is meant to be compret ive and encc ing all
closed-loop supply chain operations. The framework divides the closed-loop supply
chain into four main modules: transform to forecast (TTF), transform to demand (TTD),
retransform to forecast (RTTF), and retransform to demand (RTTD). The term
transform is used instead of ‘make’ to “avoid any implicit or explicit limitations to
manufacturing since value is created in terms of both form and place.” (Wikner & Tang,
2008: 350). For example, one could transform raw materials to components (i.e.
manufacture); or components to assembled products (i.e. assembly). Also one could
transform products from one place to another (i.e. transportation). This creates generic
modules that could be parameterized depending on the process type: manufacture,
assemble, transport, etc.
The framework allows the four different modules (i.e. TTF, TTD, RTTF, and RRTD) to
be integrated together into different configurations to model different supply chain
structures. The modules could be interchanged to represent different configurations and
simulate different scenarios.
The framework satisfies design guideline 2; each module has a specific role and is
structurally different from the other. TTF is forecast driven, and thus products are
stocked at the end of the process, also the input to the module in terms of materials is
forecast driven as well as the materials transformation rate. TTD is demand driven, and
thus there is no stock at the end of the process, also the input to the module in terms of
Modularity in System Dynamics:
i d-loop supply chain conf
materials is demand driven as well as the materials transformation rate. RTTF and
RTTD are essentially the same as TTF and TTD respectively, except that they resemble
the reverse supply chain processes. Thus they are slightly more complex as they involve
uncertainties in quality, quantity and time of returns.
These modules have similar system dynamics structures, for example: demand
forecasting, orders processing, ordering design rules, and costs and revenues. This led
us to take advantage of design guideline 4, which states that each module should have a
distinct structural skeleton. The skeleton allows for the integration of mini-modules in a
hierarchal manner. In that case, a mini-module is a module that is developed to be
integrated within the interface of the main modules. These mini-modules are repeated
structures that are used to build the main modules. For example, the TTF module can
have different forecasting strategies, and thus in the TTF skeleton you could put
different interchangeable forecasting mini-modules that represent the system being
modelled. The same mini-module could also be used in the RTTF module. In summary,
the main modules will have different structural skeletons that make use of more basic
mini-modules like: forecasting, order decision rules, orders processing, and costs and
revenues.
The modules interface should allow for material, information, and cash to flow between
modules. However, in the context of this study, we will not consider the cash flow
between modules and thus it will be disregarded. The information flows will be divided
into Outgoing Ordering Rate and Incoming Ordering Rate. And the material flows will
be divided into Material Inflow and Material Outflow. Thus each module essentially
has the interface shown in Figure 2. Other inputs to the modules could be defined, like
capacity limitation and forecasting inputs, but these inputs are not as necessary as the
ones shown Figure 2. This is because the two inputs and outputs defined in Figure 2
represent the main interaction between the modules while other inputs and outputs are
used as way to create flexibility in representing different systems, which satisfies design
guideline 4. This framework provides the design rules asked for in design guideline |
required for independent module development. And thus this framework provides the
basis for developing system dynamics modules in supply chain.
Modularity in System Dynamics:
it d-loop supply chain
Module A Module B
Outgoing Incoming Outgoing Incoming
Ordering Rate Ordering Rate ~ Ordering Rate Ordering Rate
Material Material Material Material
Inflow Outflow Inflow Outflow
Figure 2:Main interface between two modules, where the modules’ inputs are in normal text and
outputs are in bold text. The dashed lines represent information flow while the solid lines represent
material flow
We have developed four different system dynamics modules to be used for modelling
supply chain in accordance to the theoretical framework presented above. Three of the
four modules are the ones discussed above, which are TTF, TTD, and RTTD. The TTF
module would be general enough to be able to represent RTTF processes as well. In
addition to these I developed one extra version of TTF module, which is TTF-Push
module (TFF-P). The reasoning for developing this module is the fact that sometimes
material is pushed through the supply chain without ordering. For example the
European Union put forward legislations, like the WEEE directive, that forces the
companies to take care of their products end-of-life (Govindan, Soleimani, & Kannan,
2014). This makes companies responsible for collecting products from consumers after
usage, and thus the material is pushed to the supply line once the customer decides to
return the products. Also some companies decide to accept any product returns
regardless of the quality and time of return in order to increase their return volume.
After the material is returned the company can then control their stock of transformed
material by controlling the disposal rate of excess or failed products. This is represented
in the TTF-P module, which controls the stock of material by controllable disposal
rather than material ordering.
In Figure 3, we present the standard version of the TTF stock and flow structure as an
example of the modules developed. The figure shows the mini-modules used (grey
boxes), and their inputs and outputs as well as the main stock and flow structure. The
TTF module’s inputs and outputs are shown as curved squares on the module frame;
specifically, the module’s outputs are shown in bold text.
Modularity in System Dynamics:
i p supply chain confi
Transform to
Forecast
Forecasted Demand Demand
-D forecasting
loutgoing-orders|_ Desired Material
; Inf
processing nn
‘incoming
Ordering
Rate
1OR
Forecasted Material
Disposal Rate
lincoming-orders
processing
Disposal
forecasting
Material Outflow
DMO
Stock of
Material in
Material
Inflow
Start Rate
TSR
Time To Deliver
Percentage of
‘Material Disposed
PMD
Figure 3: Transform to forecast (TTF) module general structure
In Figure 3, materials could represent things like: raw materials, components, products.
livestock, or crops. The transformation process could represent things like: material
extraction, production, manufacturing, assembly, maturation, or plantation. In this
module the material goes through the transformation process then it is stocked in the
Stock of Transformed Material. This stock and flow structure is very similar to
Sterman's (2000: 710) Production-Inventory model. The Jncoming Ordering Rate input
is fed into the Incoming-Orders processing mini-module, which is responsible for
determining the Desired Material Outflow. The Stock Control mini-module processes
the Forecasted Demand, Stock of Transformed material, Stock of Material in
Transformation, and Forecasted Material Disposal Rate to give out the Desired
Material Inflow that is responsible for controlling the modules
stocks. The Desired
Material Inflow is turned into Outgoing Ordering Rate after passing through the
10
Modularity in System Dynamics:
it d-loop supply chain
Outgoing-orders processing mini-module. The Outgoing Ordering Rate is used as an
input to another module, which then gives the needed Material Inflow to supply the
stocks with the desired materials
In this module, the Material Inflow initializes the start of the transformation process;
such that once the materials arrive they are directly put in the process of transformation.
And so, the Transformation Start Rate is equal to the Material Inflow but is limited by
the Inflow Capacity limitation. The Inflow Capacity Limitation is used to represent the
capacity limitation of the transformation process. It can either be infinite, constant, or
an input from a capacity module. The Transformation Start Rate accumulates in the
Stock of Material in Transformation, while the Transformation Rate depletes the Stock
of Material in Transformation. And so, the Stock of Material in Transformation
represents the amount of material under transformation. After the material is
transformed it is stored in Stock of Transformed Material. The Material Outflow
depletes the Stock of Transformed Material and is determined by the Desired Material
Outflow but limited by the available amount of Stock of Transformed Material and Time
To Deliver as well as the Outflow Capacity Limitation. The Outflow Capacity
Limitation acts in the same way as the Inflow Capacity Limitation.
There are multiple versions of mini-modules used in the main-modules; each mini-
module is used to add a customizable function to the overall module. This leads to
multiple possible formulations of each developed module. Although this creates
flexibility in adapting the modules to different needs, it might confuse the reader. For
this reason, we have developed a naming convention that will aid the reader in
identifying which mini-modules are used within the specified module. Appendix B
shows the usage and compatibility of each mini-module developed as well as the
naming convention used in this study.
The modules developed are formulated using previously used and validated supply
chain models in system dynamics as well as qualitative descriptions of the modules’
functions. The TTF/TTD/RTTD modules’ main stock and flow structure is well
representative of a transformation process. Obviously there could be certain
transformation processes that need a more detailed stock and flow structure, but this
would not serve the purpose of this article. The modules developed have adequate detail
Modularity in System Dynamics:
i d-loop supply chain confi
and boundary to remain general yet representative to a useful degree for most
transformation processes.
We have tested the module’s structure and behaviour rigorously, within our capacity, to
eliminate the possibility of modelling errors. In that aspect, all formulations of the
modules are dimensionally consistent. Also various extreme condition tests were
passed. In addition, we have performed the boundary-adequacy test, behaviour
reproduction test, and behaviour prediction tests on all developed modules. Various
examples of the validity tests used on the modules are presented in Appendix C.
Using the modules to represent different Supply chain
configurations
Wikner and Tang (2008) identified 15 different supply chain configurations using four
distinct supply processes (i.e. TTF, TTD, RTTF, and RTTD). Table 1 shows the
different supply chain configurations, where any intersection between a column and a
row is one possible unique supply chain configuration. Out of the 15 configurations
Wikner and Tang identified nine hybrid configurations that involve an integration
between at least one forecast driven process and one demand driven process (shown as
roman numerals in Table 1). The forecast driven processes could consist of TTF and/or
RTTF while the demand driven processes could consist of TTD and/or RTTD. For
example, Configuration IX is an integration of the forecast driven processes TTF and
RTTF, and the demand driven processes TTD and RTTD.
Table 1: 15 different supply chain configurations (Wikner & Tang, 2008: 358)
Demand _ Driven
- TTD RTTD TTD+RTTD
Forecast Driven
- - TTD RTTD TTD+RTTD
TIF TIF I Il Il
RTTF RTTF IV Vv VI
TTF+RTTF TTF+RTTF VII VII Ix
Figure 4 shows how Wikner and Tang (2008) illustrate configuration IX using their
framework. The customer orders decoupling point (i.e. CODP in Figure 4) separates the
forecast driven supply from the demand driven supply. In configuration IX, customers
can purchase a product, use it, and then either return it to be repaired/serviced (i.e.
12
Modularity in System Dynamics:
it d-loop supply chain
through RTTD), return it without receiving replacement (i.e. through RTTF), or dispose
it. This configuration represents the most ‘complete and integrated’ closed loop supply
chain in comparison to the other configurations (Wikner & Tang, 2008). For example,
in configuration V, when the customer returns a product it goes through RTTD, and if
the core of the product is damaged, it is disposed to RTTF and replaced with another
core from the stock of retransformed cores. If, however, the core could not be replaced,
then the order cannot be satisfied. This is in contrast to configuration IX, where a newly
manufactured core form TTF or TTD could replace the damaged one. Wikner and Tang
(2008: 360) give the following example for configuration IX:
“In an automotive parts remanufacturing company for example,
engines cores are dismantled to obtain parts and components,
which can be further used in remanufacturing. When a customer
sends back an engine core and requests refurbishing, an RTTD
order is released, and supplementary materials are withdrawn
from the TTF and RTTF inventory at the CODP. The
refurbished engine will be sent back to the original customer. In
the second case (more often in this company), a customer
simply identifies the model of engine in his order for a
replacement, and in the meantime he will send back a core. This
will release a manufacturing order (TTD). In the assembly of an
engine, additional new material may be purchased (TTF) in
addition to the remanufactured parts and components (RTTF)
from previously returned engines.”
Forecast driven supply Demand driven supply Demand
Figure 4: Supply chain configuration 1X (Wikner & Tang, 2008: 358)
Modelling supply configurations
In this section, we use the modules to model three different configurations for a
hypothetical electronics company ‘ABC’ selling product ‘XYZ’. This is done to
illustrate the modules ability to represent these configurations, and drew valuable
13
Modularity in System Dynamics:
i d-loop supply chain conf
insights from them. The data for product XYZ (i.e. product life cycle, residence time,
manufacturing time, etc.) is estimated from Georgiadis et al.'s (2006) paper. The model
is shown in Figure 5. In this model Company ABC manufactures components for
product XYZ according to forecast (i.e. Module A). Then, product XYZ is assembled
according to demand (i.e. Module B). Products are sold and returned for repair after an
average Time to Damage. Module C receives products for repair and orders necessary
components from Module A or D according to the specified ordering ratio. The
ordering ratio is the ratio of components ordered from module A to components ordered
from module D to supplement the repair process (i.e. material needed for module C).
In this model, we investigate the effect of changing the supply chain configuration on
the total disposed products. The variable Total disposed products acts as a measure of
resource efficiency over the lifetime of products, such that when the total disposed
products are high it is an indication of loss of resources and vice versa. Changing the
supply chain configuration is done by simply varying the ordering ratio. When varying
the ordering ratio we determine if module C (which is RTTD) receives materials to
supplement the repair process from module A only (which is TTF), module D only
(which is RTTF) or both module A and D. In essence we are observing the effect of
changing the ordering ratio on total disposed products.
Modularity in System Dynamics
representing closed-loop supply chain configuration
®
Manufacturing
T
ss
MDR mo } ry
1cL=99999 |4 Sc; 99999 \\[ tet=99999
© a)
Ratio _Repair
Remanufacturing =
TTF-P 11120 RITD-200
A oor
Ty On MRI=O MO
MDR Mo Mon Mw
1ei=99999 |_| oci=99999 Oct=99999 [4 ICL=99999
Diagram annotations
<— Material flows
-<-— _ Information flows
“aeoent [seater]
Purchasing rate
Repairing rate
Time to
damage
Products in use
Repair return
rate
Products
company
Customer
disposal rate
End of life
\ time
|
Percentage
returned
1
Return Rate
End of life
time
Total
~™ disposed
products “%y
Products
disposed by
customers
Figure 5: Closed loop supply chain model in modules
Modularity in System Dynamics:
i .d-loop supply chain confi
Model formulation
Module E, shown in Figure 5, is custom made for the model. The module’s
mathematical formulation is shown in Equation set 1. The data for the model are
estimated from Georgiadis et al.'s (2006) paper and are shown in Data Table A-1 in
Appendix A. It is worth noting that we assumed that components and products retain
the same units, such that one component is enough to produce one product. This is done
for model analysis simplification.
Equation set 1: Module E in the closed loop supply chain model mathematical formulation.
ProductsInUse¢y at qd)
= ProductsInUsetsat
* (PurchasingRate + RepairingRate — RepairReturnRate
— ReturnRate — CustomerDisposalRate)
ProductsDisposedByCustomersy, gt
= ProductsDisposedByCustomers; + dt (2)
* (CustomerDisposalRate)
ProductsDisposedByTheCompany;,at
= ProductsDisposedByTheCompany; + dt * (MDRp) @)
PurchasingRate = MOg
RepairingRate = MOc (4)
RepairReturnRate = ProdutsInUse/TimeToDamage (5)
ProductsInUse a (6)
ReturnRate = Endoflifetime * (Perc urned) (7)
CustomerDi IRate = Eroaucisise «(1- C Returned)
; . ~ Endoflifetime _ (8)
TotalDisposedProucts
= ProductsDisposedByTheCompany
+ ProductsDisposedByTheCustomers (9)
Results and analysis
All the simulation runs shown below are in response to the demand shown in Figure 6.
In the first run we have set the ordering ratio to 1, meaning that all components needed
for repair are ordered from module A and non are ordered from module D. This
represents configuration III in Table 1, where the supply chain has TTF, TTD, and
RTTD processes. In this specific case, the percentage returned (i.e. the percentage of
products returned by customers after products’ end of life) is set to zero since there is
no facility to collect customers’ disposed products (i.e. no RTTF process exists).
Modularity in System Dynamics:
it .d-loop supply chain
—— Demand
1,000
Products
500
0 50 100 150 Weeks 200 250 300 350
Figure 6: Time-graph used as the demand for new products in the model
Figure 7 shows the number of products disposed by the customers and the company. In
the figure, there are approximately 340,000 products disposed in total, out of which the
customers dispose 57% of the total; the rest is disposed by remanufacturing facilities.
500 T T T T T T T T T T T
™ Products disposed by the company Products disposed by customers
400 -—
Thousand Products
w
S
6
200
100
0 +—
1 64 128 192 256 320 384 448 512 576 640 704
Weeks
Figure 7: Number of products disposed by the company and customers for the first simulation run
It is worth noting that electronics disposed by remanufacturing facilities are more likely
to be recovered faster than the ones disposed by customers. This is because companies
are more likely to acquire value by selling its waste in bulk to independent third party
remanufacturing and recycling facilities. In addition, governments, often, impose strict
regulations for manufacturers to take care of their electronics waste in an
environmentally friendly manner. The U.S. Environmental Protection Agency (EPA)
reported that individual consumers have a tendency to store unused electronics longer
than commercial consumers (U.S. Environmental Protection Agency, 2011). As an
example, on average, individual consumers send their desktop-computers to end-of-life
management facility after 12.5 years, while commercial consumers sent their desktop
computers after 4.6 years (U.S. Environmental Protection Agency, 2011). Furthermore,
the Consumer Electronics Association (CEA) argues the need to increase consumer
17
Modularity in System Dynamics:
: d-loop supply chain conf
awareness for appropriately disposing electronics after end-of-life; a survey conducted
by CEA indicated that 58% of consumers know where to discard their electronics
appropriately, and that 18% of consumers discard electronics in the trash (Consumer
Electronics Association, 2014). This indicates that it is desirable to have recycling and
remanufacturing facilities disposing products instead of individual consumers.
In the second simulation run, the ordering ratio is set to 0.5, meaning that half the
components needed for repair are ordered from module A and the other half is ordered
from module D. This represents configuration IX in Table 1, where the supply chain has
TTF, RTTF, TTD, and RTTD processes. Figure 8 shows the number of products
disposed by customers and the company when percentage returned equals 0 and 0.5.
500 r T r r r r r r T
age Products disposed by the company _ Products disposed by customers
| esrecplimcal
= ety
es
Thousand products
w
Ss
3
2.
un
aL. 64 128 192 256 320 384 448 512 576 640 704
Weeks
Figure 8: Number of products disposed by the company and customers for the second simulation run
with percentage retuned=0 or 0.5
In the figure we see that there are approximately 270,000 products disposed in total, out
of which customers dispose 72% of the total when the percentage returned=0 (top
figure) and 36% when the percentage returned=0.5 (bottom figure). These two runs
show that configuration [IX is more efficient than configuration III in terms of material
usage. Configuration IX decreased material disposal by 20% compared to configuration
III. The 20% non-disposed products in configuration IX were put back into the supply
chain and were remanufactured, thus increasing the lifetime value per manufactured
product, and increasing the company’s resource efficiency. This also goes in line with
18
Modularity in System Dynamics:
it d-loop supply chain
Wikner and Tang (2008) description of configuration IX as a complete closed loop
supply chain configuration. In the second simulation run, the percentage returned
parameter had no effect on the total amount of materials disposed. However, it
increased the percentage of materials disposed by the company instead of consumers;
this is an expected result. As customers return more products to the company, the
company disposes more products such that it is able to meet demand while minimizing
storage.
In the third simulation run, the ordering ratio is set to 0, meaning that all the
components needed for repair are ordered from module D. This represents configuration
VI in Table 1, where the supply chain has RTTF, TTD, and RTTD processes. Figure 9
shows the number of products disposed by customers and the company when
percentage returned equals 0 and 0.5.
oO ] Repair incoming orders backlog (IOB.c) | |
Percentage returned=0 ----- Percentage returned=0.5
40 —
_—
20 7
0 “= “ee
500 T 7 7 T T : , T T 7
400 _._BProducts returned by the company & Products returned by customers
300 | | |
rehurnedo || |
200
100
0 es
#500
o
5
3 400
a
B 300
5 ge returned=0.5,
2 200
=e
100
; ee
1 64 128 192 256 320 384 448 512 576 640 704
Weeks
Figure 9: Number of products disposed by the company and customers and the repair incoming orders
backlog for the third simulation run with p or 0.5
Modularity in System Dynamics:
i d-loop supply chain conf
In the figure, we see that when the percentage returned=0 there are approximately
200,000 products disposed in total, out of which customers dispose 79% of the total
(top figure). In addition, when the percentage returned=0.5 there are approximately
195,000 disposed products, out of which customers dispose 50% of the total (bottom
figure). Furthermore, when percentage returned =0, there are 50,000 incoming orders
backlog in module D (i.e. repair); this is due to the fact that there are not enough
products returned to be disassembled and remanufactured to supplement repair. This
confirms Wikner and Tang (2008) description of configuration VI, where RTTD orders
may not be satisfied if the core could not be replaced. On the other hand, when
percentage returned=0.5 the incoming orders backlog is kept at appropriate levels due
to adequate returns. This shows that such a configuration needs sufficient levels of
returns. Furthermore, since all repair components are remanufactured from used
products, the total products disposed decreased by 40% compared to configuration III.
The previous simulation runs show the performance of three different supply chain
configurations in terms of material efficiency and orders satisfaction. The simulation
results show that different configurations are suitable for different services. For
example, companies with high volume of manufacturing could take advantage of
configuration IX, to reduce their material consumption. Such configuration could also
extend the life cycle of a product by offering repair and maintenance services. This
however requires high returns, and thus companies need to increase end-of-life return
awareness of their customers. On the other hand, smaller service shops that operate as a
RTTD process could seek TTF/RTTF suppliers for new/used components that could be
used in repair and maintenance. Furthermore, companies that have configuration VI,
could offer a replacement policy through a TTD process when the core is not repairable
or replaceable. For example, the customer would bring the damaged product and
receive a new replacement from a TTD process, while the old product is sent back to a
RTTF process. This way companies will avoid high backlogged orders shown in the
third simulation run.
This model is an example of how modules could be used to represent different supply
chain configurations, and draws valuable insights from simulating different scenarios.
The model developed is in line with the conceptual descriptions that Wikner and Tang
(2008) put forward for configurations III, VI and IX. Furthermore, in this example, we
20
Modularity in System Dynamics:
it d-loop supply chain
have seen how mini-modules could be easily customised to represent the system under
study.
Conclusion and future research
The article was set out to develop system dynamics modules that can be used in
modelling supply chain systems. Modellers are meant to use the modules by plugging
them together to represent supply chain systems. The article has also sought to provide
a proof of concept regarding the applications and benefits of modularity in the system
dynamics field, specifically in modelling supply chains. There are similar attempts to
modularity in the system dynamics field; however, there is no attempt to develop
applicable system dynamics modules in the supply chain field. To pursue these goals,
we followed a systematic deductive research approach where we extracted module
development design guidelines from disciplines that have successfully applied the
concept (e.g. software development field). The design guidelines provided us with
systematic approach to modularity. We, then, used Wikner and Tang (2008) theoretical
framework to apply modularity in supply chain systems. The theoretical framework
categorized the supply chain processes to processes based on forecast and processes
based on demand, such that the Customers Orders Decoupling Point (CODP) divide the
processes based on forecast from processes based on demand. In addition, the
framework differentiates forward from reverse supply chains. As such, the theoretical
framework provides a complete set of processes that could be integrated together to
model different supply chain configurations. Finally, we have used reoccurring system
dynamics structures in supply chain modelling to develop four distinct modules. The
deductive approach was used to take advantage of the vast literature on supply chains
and provide a systematic approach to modularity in system dynamics that is relevant
and applicable to the supply chain field as well as the system dynamics field.
This study contributes to system dynamics methodology in various ways. First, system
dynamics practitioners could use the design rules specified in developing modules for
different applications. The article presents a proof of concept of the potentiality of
modularity in system dynamics. Second, the modules developed offer a tool for
efficiently modelling supply chain systems. Experienced and inexperienced modellers
could develop models to represent supply chains in less time and effort and with
minimal errors. In the study, we have shown that the modules were able to replicate
21
Modularity in System Dynamics:
i d-loop supply chain conf
previously validated models. Furthermore, we have developed an_ intuitive
understanding of the modules and their roles, which shows that when practitioners use
modules frequently they can easily identify their role in the system, and intuitively
understand their interactions within a complex system. Thus, the modules offer a tool to
articulate complex models in simple diagrams and generate intuitive understanding of
the system. Third, the modules offer a practical tool in understanding supply chain
systems. For example, the modules were used to represent three different configurations
of closed loop supply chains. Furthermore, when developing models from scratch, one
can only change parameters to test for different scenarios, leaving the modeller with
great difficulty in changing the model’s configuration or overall structure. On the other
hand, modules offer an easy way of reconfiguring the supply chain; plugging the
modules together in different configurations offers a valuable tool in scenario analysis.
This also offers a tool for adequate boundary assessment, by adding or removing
modules and comparing the model’s behaviour with the system’s behaviour. Fourth, the
innovative way of structuring the modules in a central interface with different mini-
modules, encourages cumulative learning. Modellers could build and share their mini-
modules, allowing others to use them and build on them. Thus, decreasing time and
effort while advancing the system dynamics methodology in modelling supply chains.
In addition, when modules are used in different systems, modellers could improve on
them by identifying modelling or structural errors. Thus, increasing the validity of the
modules, and their practical use. Furthermore, we imagine a great benefit for consulting
companies using system dynamics modules in modelling supply chains. The companies
could take advantage of previously developed modules and mini-modules to increase
their efficiency and validity in modelling supply chains.
In summary the article has illustrated the value of modularity in the system dynamics
field and, in particular, modelling supply chains. However, we came across some
limitations that pave the way to future research in modularity and its applicability in
system dynamics. First, the system dynamics modelling software packages are
“modularly unfriendly”. One of the most frustrating examples is that the software does
not offer input/output plugins. As such, we recommend that system dynamics software
would offer a tool in which users could develop a module, specify its inputs and
outputs, and the software would automatically show the input and output nodes, so that
the user could simply connect a module’s output node to another module’s input node.
22
Modularity i in System Dynamics:
d-loop supply chain
Lastly, the modules should be arranged in a hierarchical class library, such that when a
user changes a module class in the library, all the modules “under” this class change
accordingly. This is extremely efficient; when a modeller identifies a structural error,
he/she can change the module in the library and the software automatically changes all
the modules used under this class. The second research limitation is the restrictiveness
of the modules; it is not possible to represent all kinds of supply chain systems using
the modules developed. This is an expected consequence since some systems could be
very specific and thus difficult to represent with modules. This opens up an interesting
future research direction, in which the modules developed could be used to replicate
system dynamics models while testing their applicability on different supply chain
systems.
Bibliography
Angerhofer, B. J., & Angelides, M. C. 2000. System dynamics modelling in supply
chain h review. Simulation Conference, 2000. ...: 342-351.
http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=899737, January’ 10, 2015.
Baldwin, C. Y., & Clark, K. B. 1997. Managing in an age of modularity. Harvard
Business Review, (September 1997): 84-93.
Baldwin, C. Y., & Clark, K. B. 2000. Design Rules: The Power of Modularity Volume
J (illustrate.). MIT Press.
Consumer Electronics Association. 2014. Importance of Ecycling: CEA Study says
More People are Recycling Electronics. http://www.ce.org/News/News-
Releases/Press-Releases/2014/Importance-of-Ecycling-CEA-Study-says-More-
People, May 26, 2015.
Das, D., & Dutta, P. 2013a. A system dynamics framework for integrated reverse
supply chain with three way recovery and product exchange policy. Computers &
Industrial Engineering, 66(4): 720-733.
http://linkinghub.elsevier.com/retrieve/pii/S03608352 13003094, November 8,
2014.
Das, D., & Dutta, P. 2013b. Simulation study of an integrated reverse logistics in fuzzy
environment. Lecture Notes in Electrical Engineering.
http://www.scopus.com/inward/record.url?eid=2-s2.0-
84872521717&partnerID=40&md5=03b5e023d754164984af8f7c619907d4.
Forrester, J. W. 1961. Industrial Dynamics.
http://books.google.nl/books?id=4CgzAAAAMAAJ, Wiley.
Forrester, J. W. 1968. Market Growth as Infl d by Capital In
http://books.google.nl/books?id=BtKunQEACAAJ, Industrial Management
Review.
23
Modularity in System Dynamics:
i .d-loop supply chain confi
Forrester, J. W. J., & Senge, P. P. M. 1980. Tests for building confidence in system
dynamics models. TIMS Studies in the Management Sciences, 14: 209-228.
http://www.sdmodelbook.com/uploadedfile/32_0f93d902-b52f-4b 1 f-a658-
5a5d144c3d1b_Tests for Building Confidence JWF Senge.pdf, January 13, 2015.
Foster, I. 1995. Designing and Building Parallel Programs: Concepts and Tools for
Parallel Software Engineering.
Georgiadis, P., & Athanasiou, E. 2010. The impact of two-product joint lifecycles on
capacity planning of remanufacturing networks. European Journal of
Operational Research, 202(2): 420-433.
http://linkinghub.elsevier.com/retrieve/pii/S037722170900352X, January 9, 2015.
Georgiadis, P., & Besiou, M. 2008. Sustainability in electrical and electronic equipment
closed-loop supply chains: A System Dynamics approach. Journal of Cleaner
Production, 16(15): 1665-1678.
http://linkinghub.elsevier.com/retrieve/pii/S0959652608001091, September 9,
2014.
Georgiadis, P., Vlachos, D., & Tagaras, G. 2006. The Impact of Product Lifecycle on
Capacity Planning of Closed-Loop Supply Chains with Remanufacturing, 15(4):
514-527.
Gibb, A. G. F. 2001. Standardization and pre-assembly- distinguishing myth from
reality using case study research. Construction Management and Economics,
19(February 2015): 307-315.
Govindan, K., Soleimani, H., & Kannan, D. 2014. Reverse logistics and closed-loop
supply chain: A comprehensive review to explore the future. European Journal of
Operational Research, 240(3): 603-626.
http://linkinghub.elsevier.com/retrieve/pii/S037722 1714005633, July 17, 2014.
Hines, J. 1996. Molecules of Structure.
https://www.vensim.com/documentation/Models/Molecules/molecule.pdf.
Kamath, N. B., & Roy, R. 2007. Capacity augmentation of a supply chain for a short
lifecycle product: A system dynamics framework. European Journal of
Operational Research, 179(2): 334-351.
http://linkinghub.elsevier.com/retrieve/pii/S037722 1706002141, December 8,
2014.
Mall, R. 2009. FUNDAMENTALS OF SOFTWARE ENGINEERING.
http://books.google.no/books?id=CFhtgKrKLIEC, PHI Learning.
Olhager, J. 2003. Strategic positioning of the order penetration point. International
Journal of Production Economics, 85: 319-329.
Olhager, J. 2012. The role of decoupling points in value chain management. In H.
Jodlbauer, J. Olhager, & R. J. Schonberger (Eds.), Modeling value: 37-47.
Physica-Verlag Heidelberg.
Poles, R. 2013. System Dynamics modelling of a production and inventory system for
24
Modularity in System Dynamics:
it d-loop supply chain
remanufacturing to evaluate system improvement strategies. International
Journal of Production Economics, 144(1): 189-199.
http://linkinghub.elsevier.com/retrieve/pii/S09255273 13000704, December 24,
2014.
Richardson, G. P. 2011. Reflections on the foundations of system dynamics. System
Dynamics Review, 27(3): 219-243.
Sterman, J. 2000. Busi dj ics : systems thinkii
world. Boston: Irwin/McGraw-Hill.
and modeling for a le
Sterman, J. D. 1989. Modeling Managerial Behavior: Misperceptions of Feedback in a
Dynamic Decision Making Experiment. Management Science, 35(3): 321-339.
U.S. Environmental Protection Agency. 2011. Electronics Waste Management in the
United States Through 2009.
http://www.epa. gov/epawaste/conserve/materials/ecycling/docs/fullbaselinereport2
O11.pdf.
Vlachos, D., Georgiadis, P., & Iakovou, E. 2007. A system dynamics model for
dynamic capacity planning of remanufacturing in closed-loop supply chains.
Computers & Operations Research, 34(2): 367-394.
http://linkinghub.elsevier.com/retrieve/pii/S0305054805001000, January 8, 2015.
Wikner, J., & Tang, O. 2008. A structural framework for closed-loop supply chains.
The International Journal of Logistics Management, 19(3): 344-366.
http://www.emeraldinsight.com/doi/abs/10.1108/09574090810919198, January 9,
2015.
Wolstenholme, E. F., & Coyle, R. G. 1983. The development of system dynamics as a
methodology for system description and qualitative analysis. Journal of the
Operational Research Society, 34(7): 569-581.
25
