System Dynamic Modeling

Operations Research
Term Paper
Group 7
Anuradha Dhote (PGP/17/010)
Jignesh Rathod (PGP/17/037)
Sushant Kumar (PGP/17/053)
Usha Bhakuni (PGP/17/057)
Yashasvi Kansotia (PGP/17/062)

P a g e | 2

1. Abstract
System Dynamic Modelling is an approach to frame and solve complex problems. It is widely used
for policy design and analysis. System dynamics analysis is a part of Systems Theory which is used to
study the dynamic behaviour of complex systems. The theory is based on acknowledgement of the
fact that the inter-relations between the various components of a model, time delays between
processes are as important to analyse the behaviour of a model as is the analysis of individual
components. SD models solve the problem of simultaneity (mutual causation) by updating all
variables in small time increments with positive and negative feedbacks and time delays structuring
the interactions and control.
In this paper, we have discussed the basic methodology of the technique, and the modelling process.
Then, we have discussed the applications of SDM ranging from conventional ones like project
management and science and engineering to exploring the concepts of SDM in Brand management
and public health issues.
2. Historical background
This technique of solving problems was developed in 1950s by Prof Jay Forrester of Massachusetts
Institute of Technology, during his involvement with General Electric. He developed his insights
about the foundations of engineering which helped him in the creation of system dynamics. He
studied the problem of instability in GE employment. In this he did hand simulations (or calculations)
of the stock-flow-feedback structure of the GE plants, which included the existing corporate
decision-making structure for hiring and layoffs. Forrester was able to show how the instability in GE
employment was due to the internal structure of the firm and not to an external force such as the
business cycle. These hand simulations were the beginning of the field of system dynamics.
During 1960s, the problem solving evolved from hand simulations to formal computer modelling.
During this time programming languages based on this approach were developed, like SIMPLE and
DYNAMO.

3. Why is System Dynamic Modelling needed
A model usually has many factors related to it and it is very important to study each of these
factors to gain complete knowledge about the subject, though it is not so easy, it takes a lot
of effort to integrate these and study the changes in one relative to the other, this is exactly
where SDM comes to our rescue, provides us with a common platform and thus it enablesus
to combine the various factors involved in a model and study the whole model as one single
unit rather than studying different factors one at a time.
In the long history of evolution it has not been necessary until very recent historical times
for people to understand complex feedback systems. Evolutionary processes have not given
us the mental ability to interpret properly the dynamic behaviour of those complex systems
in which we are now embedded. - Forrester, 1973
P a g e | 3

Another main reason why we need SDM is that we now have to face Complex systems and if
we need to make an improvement to the same one cannot change one part without a
certain risk of affecting other remote parts. Thus we are obliged to understand the whole
system.


4. System Dynamic Modelling Approach
System Dynamic Modelling provides an approach to analyse complex real world problems over a
period of time. In this approach, the modeler attempts to identify the patterns of behaviour of the
system measured by certain parameters of the system, and then build a model that can mimic the
patterns. Once the model is built, it can be used for testing by altering a systems behaviour.
Time paths
A critical step in examining a system is to identify its key patterns of behavior -called "time paths."
They describe how a system behaves over time. There are various types of time paths that a system
follows linear, exponential, S-shaped, Oscillatory, etc.
Stocks and Flows
System Dynamic Modelling is based on the Principle of Accumulation which states that all dynamic
behaviour in a system occurs when flows accumulate in stocks. The stock-flow structure is the
simplest dynamic system. In case of SDM, information and non-information move through flows and
are stored/ processed in stocks.
Theoretically, a stock can have infinite number of inflows as well as outflows. However, in practice
only 5-6 flows are present with a stock.
Stocks have the following four characteristics
a) Have memory
b) Change the time shape of flows
c) Decouple flows,
d) Create delays.
These characteristics are crucial in determining the behaviour of the system.
Feedback
Systems can be classified as being open or closed. In open systems, output has no influence on the
system or the inputs. Inputs will be independent of the output. Whereas, in case of closed systems,
the outputs respond to inputs and also influence the inputs. Most commonly existing systems are
closed loop systems.
Stocks and flows form a part of feedback loop that determines the relation between these
components. Every feedback loop must have one stock element.
There are two types of feedback loops
P a g e | 4

1. Positive feedback In a positive feedback loop, a change in any of the components is
amplified, and it causes a large self-change in the original direction. They are responsible for
the growth in the system.
2. Negative feedback A feedback loop that tends to diminish an initial change in any of the
variables is called a negative loop. These loops are aimed at keeping the system in a desired
state. These loops stabilize the system.
Non linearity
A system is not perfect and its limits are set by the non-linear relationships that exist between its
components. These non-linear relationships can cause change in the overall response of the system.
These relationships can cause the system to behave in a manner of an entirely different system over
time.

5. System dynamic modelling process
The modelling process consists of the following steps
1. Identify the problem In a systems dynamic problem, modelling the entire system is very
complex and its easier to focus on only one problem. It sets a boundary and helps in
considering only the relevant variables rather than the entire system. It is better to begin
with a clearly defined problem statement.
2. Develop hypothesis Once the problem statement is defined, the next step is to develop a
hypothesis about the possible cause of the given behaviour of the system. Causal loop
diagrams are used to list out some of the assumptions about the possible reasons. The
following process can be used to formulate assumptions
a. List the relevant system variables
b. Link the variables according to their inter-relationships keeping in mind the problem
statement
c. Form the feedback loops and try to analyse the possible causes.
This is an important stage to collect the information about the problem and the system.
3. Test hypothesis In this stage the model is converted into mathematical model and a
computer model is made. Then simulations are run on the computer model to see whether
the system behaves in the desired manner. If it behaves differently, then there is a problem
with either the model or the hypothesis. Then, we need to re-look the entire model.
4. Test policies Once the model has been constructed, the root cause of the problem can be
easily identified. Accordingly, solutions to the problem can be proposed. This can help in
determining the uncertainties that are present in the system and how to overcome.
Different decisions can be tested on the model and the model can be improved upon
accordingly.

P a g e | 5

Management Flight Simulator
After the model has been developed, in order to make decision making easier and explain the
working of the model to the management, a simulation is developed. This simulation is called the
Management Flight Simulator; it must have the following information
a) Decision making tools Keys to help the management take decisions about the process
b) Reports To provide information about the state of the system
c) Graphs Time series graphs allow the user to study system trends
d) Initial conditions These are used to set assumptions about the initial states of the system
parameters.
Thus, the flight simulator acts as a laboratory in which management strategies could be tested.

6. Applications of SDM
SDM has found a wide range of application in population, ecological and economic systems. Its
application helps to gain quantitative insight into the system, analyze and compare mental models
and help incorporate systems thinking into our perspective. While SDM is used to model system for
well-known applications like project management and engineering, it is also being used in areas like
brand management, hence, helping to introduce the perspective of dynamic behaviour, time paths
and causal loops in them, hence taking these newfound applications to a new level of granularity.
Let us look at a few applications which can be modelled using SDM approach -
1. Science and engineering: Physical modelling has an important place in science and
engineering. Topics like Newtonian mechanics, control systems, finite element modelling,
and chemical engineering use time-dynamic concepts, involved feedback loops and flow of
information between different points. SDM is simple, powerful, useful and natural and is
suited to modelling the subject of dynamical systems in physical science and engineering.

2. Application of system dynamics to brand management: Dynamic issues are typically not
addressed in conventional brand management approaches, for e.g. how cost cutting may
undermine long-term performance, service qualitys effect on price acceptance, intensity of
presence influencing loyalty etc. Using SDM, causal loops can be developed for how a
business operates. Taking the example of a n ethnic eating outlet, the cause-effect
relationship between increase in restaurants offering ethnic food concepts, resulting in
stimulated growth in ethnic spending and increased ethnic spending leading to increased
investment in ethnic food concepts can be modelled as an SDM causal loops

3. Project Management: Project management in various areas, whether they are construction,
software, power generation or defence et al, suffers from numerous problems, primarily,
cost overruns and delays. Customer design changes and errors when discovered towards the
end of the project, cause overtime, schedule slippage and costly overwork. This leads to
lower revenues, quality, loss of market share and differences between customers and
P a g e | 6

vendors over responsibility for overruns and costs.

Projects are complex dynamic systems with certain characteristics, making system dynamic
modelling as a suitable framework to be applied to their management and execution. These
characteristics can be explained as follows

a. Extremely complex and multiple interdependent components
b. Highly dynamic
c. Multiple feedback processes
d. Involve non-linear relationships
e. Involve both hard and soft data




















P a g e | 7

Example: System Dynamic Modelling applied to public health issues
Let us take the example of system dynamic modelling applied to public health. Below figure shows a
broader view of population health using SDM approach.


When applied to the issue of chronic disease prevention, an SDM takes into account disease
outcomes, health and risk behaviours, environmental factors and health related delivery systems.
This model explains how a hypothetical set of people affected by chronic diseases may be affected
by two types of prevention - upstream prevention of disease onset and downstream prevention of
disease complication.
1. We model the essential causal structure and policy inputs as below. There is gradually
changing net accumulation of two flows: inflow of disease onset and outflow of deaths

P a g e | 8

2. Here we map the simulation output over 50 years, in three scenarios (Status Quo, More
complications prevention and More Onset Prevention), for four variables
a. Onset prevention fraction
b. Complication Prevention fraction
c. People with disease
d. Deaths from complications
Here we see that with the increase in new tools for complication prevention lead to an increase in
preventable fraction of complications from 25% to 50%. But reduction in deaths means longer
average stay in diseased population stock, which leads to greater disease prevalence. This in turn
increases the need for resource for complication prevention. Hence, this is a vicious cycle which
health care system may suffer from given its limited resources for disease prevention and need for
best possible management of existing disease.


References
1. http://www.systemdynamics.org/DL-IntroSysDyn/
2. SYSTEM DYNAMICSMODELINGINSCIENCEAND ENGINEERING, Hans U. Fuchs
3. System Dynamic modelling for Prjoect Management, John D Sterman
4. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1470525/
5. Application of system dynamics to brand management, Vanguard Brand Management