Lesson 6 – Stock and Flow Diagrams

IE2141 Systems Thinking and Dynamics

Dr. LI Haobin, Senior Lecturer

Department of Industrial Systems Engineering and Management (ISEM)
College of Design and Engineering
National University of Singapore

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore
 6.1 – Introduction to SFDs

Lesson 6 – Stock and Flow Diagrams

IE2141 Systems Thinking and Dynamics

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 2
 Recap on CLDs

Do they have the

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore same meaning? 3
 Why SFD?
▪ Although a CLD expresses how one variable is related to another, it
does not provide enough detail to calculate the value of one variable
given the values of the others.
▪ In contrast, an SFD is a calculable representation of the system.
▪ For this to be possible, each variable must be defined and assigned
the correct units, and often new variables must be added as well.

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 4
 What is SFD?
▪ Stock and flow diagrams (SFDs) are ways of representing the structure of
a system with more detailed information, i.e., quantitative values and
equations, than is shown in a causal loop diagram (CLDs).
▪ Stocks are fundamental to generating behavior in a system;
▪ Flows cause stocks to change.

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 5
 SFD Notations

▪ Overview of a Connectors

Stock-Flow Diagram. Converters

Stock
Source Sink

In-Flow Out-Flow

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 6
 SFD Notations

▪ Sources and Sinks represent everything outside the system

Source

▪ "Sources" and "sinks" are always associated with "flow" arrows,

representing external "stocks" that lie outside the scope of study.
Sink
▪ The "source" is not an input, nor is the "sink" an output.
▪ In mathematics, no variables defined for "sources" and "sinks".

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 7
 SFD Notations
▪ Stocks are quantities that can accumulate or decrease in numbers.
▪ Flows show the movement and captures rate that quantities are
transferred.
Stock Flow
It is not a pipe,
zero space in flow

▪ The value of flows can be directly controlled

▪ The value of stocks only change when connected to an inflow
or outflow

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 8
 SFD Notations
▪ Converters hold information about the system.
▪ They store equations or constants, and affect rate of flows or another
converters (Variables/Parameters). 𝑥2 = 3

𝑥1 = 2 𝑦 = 𝑥1 − 𝑥2

Converters
▪ By default, system variables that do not have
dynamic "causal relationships" with others can be
represented by "converters“
▪ Each converter can be seen as a static subsystem

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 9
 SFD Notations Causal relationship
in CLD

▪ Connectors move information from one element to another e.g.

connects population to births.
Influence via
Connectors

𝑥2 = 3 Stocks Stocks

Flows Flows
Connector 𝑥1 = 2 Converters Converters

𝑦 = 𝑥1 − 𝑥2 Stock can only be influenced by

Flows via inflows and/or outflows.
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 10
 How does SFD complement CLD?
▪ SFD allows non-dynamic causal relationships between system
variables to be annotated on the diagram through mathematical
formulas.
▪ SFD depicts the dynamic causal relationships between system
variables through the standard forms of "stock" and "flow", with
implicit integral equations.

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 11
 How does SFD complement CLD?
▪ CLDs quickly conceptualise and clarify the overall structure, while SFDs
confirm the numerical relationships between causal factors
▪ We typically begin by drawing the CLD and then use the SFD to explore
specific dynamic causal relationships that require attention.
▪ However, in drawing SFDs we may discover additional variables or new
causal relationships, prompting us to modify the CLD accordingly.

Causal relationships in CLD Causal relationships

in SFD

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 12
 6.2 – Conversion between
CLDs and SFDs
Lesson 6 – Stock and Flow Diagrams
IE2141 Systems Thinking and Dynamics

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 14
 Mathematics in SFD CLD:

A particular type
CLD:

𝑥2 = 3

SFD:
𝑥1 = 2
Integral Equation:
𝑡
𝑦 = 𝑥1 − 𝑥2
Stock 𝑡 = න Inflow 𝑠 − Outflow 𝑠 𝑑𝑠 + Stock 𝑡0
𝑡0
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 15
 Mathematics in SFD

Stock 𝑡 = න Inflow 𝑠 − Outflow 𝑠 𝑑𝑠 + Stock 𝑡0

𝑡0

𝑋 𝑡−𝛿 𝑌 𝑡
SYSTEM
Inflow(𝑡 − 𝛿) / Stock(t)
Outflow(𝑡 − 𝛿)
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 16
 Consistency in Units Unit is consistent
through the equation

▪ In a SD simulation model represented

(unitless)
by the SFD, units for Stocks, Flows, and
(Fish) (unitless)
Converters shall be consistent, while
the equations at each Converters are
taken into consideration.

Same Unit Same Unit Same Unit

(Fish/Month) (Fish/Month)

(Fish)

Same Unit / Same Unit /

Time Time Unit is consistent Unit is consistent
through the equation (Per Month) (Month)
through the equation
Tip: If necessary, add additional coefficient in equation as
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore a unit converter, but make sure it has a real world meaning 17
 Consistency in Causal Relationship
▪ CLD is usually an aggregated version of the model represented by SFD, addressing
important variables. Mapping between CLD and SFD are described as follows.
Extension with
additional information
CLD SFD
Aggregate/simplify

Stock
Variable Flow

Converter

Connector Same arrow direction, and polarity (+/-)

based on the explicit equation
Causal
Relationship Stock-Flow In general, Inflow has an +ve impact on the stock,
Relationship whereas the outflow has an -ve impact on the stock
𝑡 (Implicit equation)
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore Stock 𝑡 = න Inflow 𝑠 − Outflow 𝑠 𝑑𝑠 + Stock 𝑡0 18
𝑡0
 6.3 – Examples of SFDs from CLDs

Lesson 6 – Stock and Flow Diagrams

IE2141 Systems Thinking and Dynamics

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 20
 Variables with the Same Unit
▪ Cases where variables A and B in CLD forming a positive
causal relationship have the same unit
▪ How to identify possible stock-flow structures?

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 21
 Variables with the Same Unit
▪ Case 1 – A and B are stocks that form a “material flow”
▪ Flow F connecting from A to B is created in SFD

Transform Aggregate

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 22
 Variables with the Same Unit
▪ Case 2 – A and B are stocks that form a “material flow”
▪ Flow F connecting from B to A is created in SFD

Transform Aggregate

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 23
 Variables with the Same Unit
▪ Case 3 – A and B are stocks that form an “information flow”
▪ Flow F connecting to B is created in SFD

Transform Aggregate

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 24
 Variables with the Same Unit
▪ Case 4 – A and B are stocks that form an “information flow”
▪ Flow F connecting from B is created in SFD

Transform Aggregate

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 25
 Variables with the Same Unit
▪ Cases where variables A and B in CLD forming a negative
causal relationship have the same unit
▪ How to identify possible stock-flow structures?

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 26
 Variables with the Same Unit
▪ Case 5 – A and B are stocks that form a “material flow”
▪ Flow F connecting from B to A is created in SFD

Transform Aggregate

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 27
 Variables with the Same Unit
▪ Case 6 – A and B are stocks that form a “material flow”
▪ Flow F connecting from A to B is created in SFD

Transform Aggregate

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 28
 Variables with the Same Unit
▪ Case 7 – A and B are stocks that form an “information flow”
▪ Flow F connecting to B is created in SFD

Transform Aggregate

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 29
 Variables with the Same Unit
▪ Case 8 – A and B are stocks that form an “information flow”
▪ Flow F connecting from B is created in SFD

Transform Aggregate

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 30
 Variables with Units of Increasing Orders
▪ Cases where variables A and B in CLD forming a positive
causal relationship have the units of increasing integral orders
▪ How to identify possible stock-flow structures?

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 31
 Variables with Units of Increasing Orders
▪ Case 9 – A is the inflow of Stock B

Transform

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 32
 Variables with Units of Increasing Orders
▪ Case 10 – A has a negative influence on the outflow of Stock B
▪ Outflow A’ from B is created in SFD

Transform Aggregate

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 33
 Variables with Units of Increasing Orders
▪ Cases where variables A and B in CLD forming a negative
causal relationship have the units of increasing integral orders
▪ How to identify possible stock-flow structures?

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 34
 Variables with Units of Increasing Orders
▪ Case 11 – A is the outflow of Stock B

Transform

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 35
 Variables with Units of Increasing Orders
▪ Case 12 – A has a negative influence on the inflow of Stock B
▪ Inflow A’ to B is created in SFD

Transform Aggregate

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 36
 Variables with Units of Increasing Orders
▪ Cases where variables A, B and C in CLD forming a causal
chain relationship have the units of increasing integral orders
▪ How to identify possible stock-flow structures?

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 37
 Variables with Units of Increasing Orders
▪ Case 13 – A is the inflow of Stock B, which has a positive
influence on the outflow of Stock B
▪ Inflow B’ from C is created in SFD

Transform Aggregate

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 38
 Variables with Units of Increasing Orders
▪ Case 14 – A is the inflow of Stock B, which has a positive
influence on the outflow of Stock B
▪ Inflow B’ from C is created in SFD

Transform Aggregate

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 39
 CLD vs. SFD for General BOTGs
▪ Cases where the structure is for Constant Behaviour

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 40
 CLD vs. SFD for General BOTGs
▪ Cases where the structure is for Exponential Behaviour

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 41
 CLD vs. SFD for General BOTGs
▪ Cases where the structure is for Goal-Seeking Behaviour

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 42
 CLD vs. SFD for General BOTGs
▪ Cases where the structure is for Oscillation Behaviour

:
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 43
 CLD vs. SFD for General BOTGs
▪ Cases where the structure is for S-Shape / Overshoot

:
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 44
 Other Examples

▪ SFD conversion from the perspective of time delays.

▪ Eight System Archetypes in Lesson 7 and Lesson 8

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 45
 6.4 – SFDs for Time Delays

Lesson 6 – Stock and Flow Diagrams

IE2141 Systems Thinking and Dynamics

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 47
 What is a Delay? 𝑋 𝑡−𝛿 Process 𝑌 𝑡
(sub-system)

▪ Process where output lags behind its input in some fashion

▪ Common in a Dynamic & Causal System
▪ Delay always contain stock-flow relationship

Stock 𝑡 = න Inflow 𝑠 − Outflow 𝑠 𝑑𝑠 + Stock 𝑡0

𝑡0
Depends on what happed in the past
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 48
 Material Delay on Causal Relationship
▪ Cases where a Material Delay occurs on a positive causal relationship
from A to B, which have the same unit
▪ How to identify possible stock-flow structures?

Input

Average Length
of Delays Delayed
Output
Infinite -Order Delay
(Deterministic)

Material Delay
and Order

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 49
 Material Delay on Causal Relationship
▪ Case 15 – A and B are input and output of a 1st order Material
Delay, with expected delay time T
▪ Transition stock S is created in SFD

Transform Aggregate

𝑆
𝐵=
𝑇

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 50
 Material Delay on Causal Relationship
▪ Case 16 – A and B are input and output of a higher order Material Delay,
with expected delay time T
▪ Transition stock S1, S2, S3 and intermediate flows F1 and F2 are created
in SFD

3×𝑆
𝐹1 =
𝑇

Transform Aggregate
3×𝑆
𝐹2 =
𝑇

3 × 𝑆3
𝐵=
𝑇

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 51
 Material Delay on Causal Relationship
▪ Case 17 – A and B are input and output of a higher order Material Delay,
with expected delay time T
▪ Alternatively, transition stock S is created in SFD, with adoption of DelayN
function

𝐵 = 𝐷𝑒𝑙𝑎𝑦𝑁 𝐴, 𝑇, 𝑁

Transform Aggregate

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 52
 Material Delay on Causal Relationship
▪ Cases where a Material Delay occurs on a positive causal
relationship from A to B, which have different units
▪ How to identify possible stock-flow structures?
Case 18

(Unit-1) (Unit-1) (Unit-2)

Transform

(Unit-1) (Unit-2)

(Unit-1) (Unit-2) (Unit-2)

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore Apply Cases 15-17 53
 Material Delay on Causal Relationship
▪ Cases where a Material Delay occurs on a negative causal
relationship from A to B
▪ How to identify possible stock-flow structures?
Case 19

Transform

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore Apply Cases 15-18 54
 Information Delay on Causal Relationship

▪ Cases where an Information Delay occurs on a positive causal

relationship from A to B
▪ How to identify possible stock-flow structures?

In Material Delays, Input / Output

are modeled as Flows
In Information Delays,
Output are modeled as Stock
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 55
 Information Delay on Causal Relationship

▪ Case 20 – A and B are input and output of a 1st order Information

Delay, with expected delay time T
▪ Bi-directional flow F is created in SFD, to adjust stock B based on A

Transform Aggregate

𝐴−𝐵
𝐹=
𝑇

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 56
 Information Delay on Causal Relationship

▪ Case 21 – A and B are input and output of a higher order Information

Delay, with expected delay time T
▪ Bi-directional flow F is created in SFD, to adjust stock B based on A

Transform Aggregate
3 𝐴 − 𝑆1
𝐹1 =
𝑇

3 𝑆1 − 𝑆2
𝐹2 =
𝑇

3 𝑆2 − 𝐵
𝐹3 =
𝑇
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 57
 Information Delay on Causal Relationship

▪ Case 22 – A and B are input and output of a higher order Information

Delay, with expected delay time T
▪ Alternatively, SMTHN function is adopted without explicit stock-flow
structure

Transform
𝐵 = 𝑆𝑚𝑡ℎ𝑁 𝐴, 𝑇, 𝑁

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 58
 Information Delay on Causal Relationship

▪ Cases where an Information Delay occurs on a negative causal

relationship from A to B
▪ How to identify possible stock-flow structures?
Case 23

Transform

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore Apply Cases 20-22 59
 Terminologies about Delay Types
▪ Physical Delays:
Higher-order material delays
▪ Transactional Delays:
Aggregation of many material / information delays
▪ Informational Delays:
Higher-order information delays
▪ Perceptual Delays:
Low-order information delays
▪ Pipeline Delays:
Highest-order (𝑛 → ∞) material / information delays Example of an
▪ Aging Chains: Aging Chain
Disaggregation of material delays into multiple stages

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 60
 6.5 – Input Parameters and Equations

Lesson 6 – Stock and Flow Diagrams

IE2141 Systems Thinking and Dynamics

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 62
 Fitting of Input Parameters
▪ Besides static decision variables, which are to be controlled by the
decision makers, all other independent variables, i.e., input parameters,
shall be adjusted to best fit a set of observed data from reality
▪ In System Dynamics model of IE2141, only deterministic parameters are
required. I.e., assume all input parameters are constant.
▪ However, in advanced system modelling,
Independent
parameters can be stochastic Variables

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 63
 Types of Parameters
▪ Count
▪ Example: size of population
▪ Average / Expectation
(across the population or over the time)
▪ Example: average lifespan, expected time delay, average number of contacts, birth rate
▪ Proportion / Probability / Ratio
▪ Example: probability of infection upon contact, proportion of resource to allocate to A,
utilization of resource
▪ Quantile / Percentile / Minimum / Maximum
▪ Example: waiting time of 95% of customers, minimum level of inventory

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 64
 Data Sources
▪ Value of Parameters shall be concluded from data observation
▪ Data Sources
▪ Actual Observation
▪ Survey
▪ Expert Opinions
▪ Library / Literature
▪ Database / Public Domain Statistics
▪ Note that, there are other roles of data in system dynamics
modelling, besides fitting Input Parameters
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 65
 Data Estimation – Graphical
▪ Plot of number of people based on age
▪ How do you estimate the number of babies below 2 years old?

Equation:

Lumped Method
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 66
 Data Estimation – Division
▪ Find the volume of a dollar bill
▪ How to find Height which is very thin?
▪ Estimate it by measuring a stack of papers and dividing it by the
number of papers.

Volume = Length x Width x Height

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 67
 Data Estimation – Scale Up
▪ How many bed pillows can fit in the back of a tractor trailer?

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 68
 Data Estimation – Aggregation
▪ Estimate the peak capacity of a MRT Train at peak hour.
▪ Assuming there are 20 train carriages with each carrying 150
passengers at peak hour and 1 train every 10 mins

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 69
 Data Estimation – Extrapolation
▪ How much fuel is burnt by Texas A&M students for all the
Thanksgiving visits?

Fuel Used
Number of Students

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 70
 Data Estimation – Analogies
▪ Estimate the volume of an average-sized man.

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 71
 Data Estimation – Geometry
▪ Estimate the surface area of an average sized man

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 72
 Equations to Quantify Causal Relationship

▪ Confirm “Causality” rather than “Correlation”

▪ Only when the cause relationship can be confirmed, e.g.,
through experiments, literature, expert opinions, interviews, etc.
▪ Data can be used to derive mathematical equations to quantify
the causal relationship, i.e., the equation can be used to
calculate the “effect” based on the “causes”.

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 73
 Possible Equations for Causal Relationship?

Variable 2
𝑦 = 𝛼 + 𝛽𝑥

𝑦 = 𝛽0 + 𝛽1 𝑥 + 𝛽2 𝑥 2 + ⋯
𝑥 𝑦
𝑦 = 𝛾0 + 𝛾1 2𝑥 + 𝛾2 3𝑥 + ⋯

Variable 1

Source: https://www.geogebra.org/m/P58s2T58

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 74
 Simple Linear Regression

▪ Consider the simple model function

𝑦 = 𝛼 + 𝛽𝑥

▪ Each data point 𝑖 can be explained by

𝑦𝑖 = 𝛼 + 𝛽𝑥𝑖 + 𝜀𝑖

▪ We aim to find 𝛼ො and 𝛽መ that minimize

σ𝑛𝑖=1 𝜀𝑖Ƹ 2
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 75
 Multiple Regression Analysis
Note that 𝑋𝑘 could also be a nonlinear form of the
1
cause variable, i.e., 𝑋𝑘 = 𝑥 or 𝑋𝑘 = 𝑥 2 , or 𝑋𝑘 = 3𝑥
▪ Consider a general model function
𝑦 = 𝛽0 + 𝛽1 𝑋1 + 𝛽2 𝑋2 + ⋯ + 𝛽𝑝 𝑋𝑝
▪ Each data point 𝑖 can be explained by
𝑦𝑖 = 𝛽0 + 𝛽1 𝑋𝑖1 + 𝛽2 𝑋𝑖2 + ⋯ 𝛽𝑖𝑝 𝑋𝑖𝑝 + 𝜀𝑖

▪ We aim to find 𝛽መ0 , 𝛽መ1 , …, 𝛽መ𝑝 that minimize

σ𝑛𝑖=1 𝜀𝑖Ƹ 2
IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 76
 Steps to Formulate an Equation
1. Focus on one Converter in the SFD
2. List data points for features (i.e., causes), and response (i.e., effect)
3. Make a guess on the type of equation, based on the observation
and plot of the function
4. Search for values of equation parameters that lead to the minimum
σ𝑛𝑖=1 𝜀𝑖Ƹ 2
▪ Use Excel for the regression analysis
https://www.youtube.com/watch?v=M1tVwND8T1o

https://www.youtube.com/watch?v=M1tVwND8T1o

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 77
 Requirement for Systems Thinking
▪ However, Regression Analysis in IE2141 Project is NOT
compulsory
▪ In Systems Thinking,
▪ “Breath” is more important than “Depth”.
▪ It is more critical to elaborate the structure of the overall system, and
identify the big causal feedback loop, before zooming into the
accuracy of data analysis for individual equations

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 78
 6.6 – Implement System Dynamics
Simulation
Lesson 6 – Stock and Flow Diagrams
IE2141 Systems Thinking and Dynamics

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 80
 Software – Stella Architect
STELLA (short for Systems Thinking, Experimental Learning Laboratory with Animation; also marketed
as iThink) is a visual programming language for system dynamics modeling introduced by Barry
Richmond in 1985. The program, distributed by isee systems (formerly High Performance Systems)
allows users to run models created as graphical representations of a system using four fundamental
building blocks. STELLA has been used in academia as a teaching tool and has been utilized in a variety
of research and business applications.

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 81
 Software – Stella Architect
Stella Architect is the definitive modeling tool to create professional simulations and presentations.
Seamlessly create, design, and publish models to share with anyone, any where, anytime. It allows you
to build rich flight simulators and learning environments on top of Stella models that clearly explain model
concepts and invite users to explore different assumptions and see the outcomes over time. Stella
Architect interfaces can be published to the web, to the desktop, or to mobile devices. With a sample
click, people can engage with your simulations anywhere on any device with a modern web browser.

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 82
 Integration Methods
▪ System dynamics models are formulated in continuous time
and solved by numerical integration.
▪ Delta Time (DT) yield an approximation of the underlying
continuous dynamics.
▪ DT is not a feature of reality
– decision rules and models
should not depend on it.

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 83
 Integration Methods
▪ Runge-Kutta (RK) Methods
https://en.wikipedia.org/wiki/Runge%E2%80%93Kutta_methods

▪ Euler Method
https://en.wikipedia.org/wiki/Euler_method

▪ Heun's method
https://en.wikipedia.org/wiki/Heun%27s_method

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 84
 Integration Methods
▪ Example: Workforce Goal-Seeking (Varying DT from 3 to 0.2)

DT = 3

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 85
 Integration Methods
▪ Example: Workforce Goal-Seeking (Varying DT from 3 to 0.2)

DT = 0.2

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 86
 Integration Methods
▪ The results of your models should not be
sensitive to the choice of time step or integration
method.
What delta time to use?
Start with a particular DT
Yes

Reduce DT by half Is result sensitive No

Accept DT / Alternate
and compare results to the change between different
integration methods

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 87
 Let's protect the environment
Lecture notes are subject to change before
the class for further improvement

THANK YOU
The preparation of this module has been made possible by
the support from IE2141 teaching team, and the past
teaching materials developed by A/Prof. Aaron Chia.

IE2141 Systems Thinking and Dynamics – ISEM Department, National University of Singapore 88