Modeling and Simulation

MODULE 2
DISCRETE EVENT SIMULATION
 SUBTOPIC 1
STATISTICAL CONCEPTS
FOR DES
• Demonstrate understanding of the several statistical concepts necessary for
developing and using discrete event simulation.
• Identify statistical methods associated with including randomness as part of a
system model.
• Identify methods associated with processing the simulation output to assess
system performance.
• Recall properties of sets and functions that are used to define a probability space.
• Investigate some of the input data modeling associated with including randomness
as part of a system model.
• Investigate some of the output data analysis associated with processing the
simulation output to assess system performance.
• Probability is a fundamental building block of
statistics.
• Statistics is the source of several very powerful
techniques used to:
❖ represent randomness
❖ generate the random variate streams that drive a
stochastic simulation
❖ interpret the output data obtained from
conducting a stochastic simulation experiment.
Set – collection of objects and usually denoted by capital letters.

Element – objects of the set (  )

Null or Empty Set – a set having no elements. (  or { } )

Examples: S = { x, y, z } xS
A={}
Note:
• There are no repeated occurrences of an element in the definition of
a set.
• An object is either an element of a set or it is not.
Set Characteristics

1) A set is said to be countable if the elements of the set can be put in a one-to-
one correspondence with the positive integers,
2) A set is called finite if it is empty or if it has elements that can counted with the
counting process terminating. A set that is not finite is called infinite.
3) If every element of set A is also an element of set B, then A is said to be
contained in B and A is called a subset of B denoted as A  B.
4) Two sets are said to be disjoint if they have no common elements.
5) The largest or all-encompassing set of elements under discussion in a given
situation is called the universal set denoted by U.
6) Two sets A and B are said to be equal, denoted A = B, if set A and set B
contained exactly the same elements.
Set Operations

1) The union of two sets A and B, denoted A  B, is the set consisting of all
elements that are contained either in set A or in set B.
2) The intersection of two sets A and B, denoted A  B, is the set consisting of all
elements that are contained in both sets A and B.
3) The difference of two sets A and B, denoted A – B, is the set consisting of the
elements contained in set A that are not contained in set B. It should be
observed that, in general, A – B  B – A.
4) Let U be the universal set and let A be a subset of the universal set. The
complement of set A, denoted by AC, is the set U – A.
Example:

Let U = { 2, 4, 6, 8, 10, 12 }
A = { 2, 4, 10 }
B = { 4, 6, 8, 10 }

What is A  B? _____________ What is A – B ? ________

What is A  B? _____________ What is B – A ? ________

What is AC ? ________
Function
• A mathematical construct consisting of three components:
1) A set of elements called the domain set X.
2) A set of elements called the codomain set Y.
3) A rule of correspondence f that associates each element of
the domain set with exactly one element of the codomain set.

❖ For x  X and y  Y, a function is denoted by f:X → Y or f(x) = y.

❖ Two functions are equal if they have the same domain set, the
same codomain set and the same rule of correspondence.
Properties of Functions

1) The range of a function f is the set of elements in the codomain Y

that are associated with some element of the domain X under the
rule of correspondence f.
2) If the codomain of a function and the range of the function are
equal, then the function is said to onto.
3) If each element of the domain X maps to a unique element of the
codomain Y, then the function is said to be one-to-one.
4) A function f has an inverse, denoted f-1, if and only if f is one-to-
one and onto.
5) For function f:X → Y one-to-one and onto, the inverse is a function
f-1:Y → X such that f-1[f(x)] = x for all x  X.
Example:

Let R denote set of real numbers, domain X = R, codomain Y = R, rule of correspondence

y = f(x) = mx + b where m and b are real-valued constants.
Graph:

For m  0, the function is one-to-one and onto.

Thus, it has an inverse function: x = f-1(y) = (1/m)y – (b/m).

For m = 0, the function becomes y = f(x) = b. All points in the domain map to a single point in the
codomain. Range of function is the set {b}. Hence, function is neither one-to-one nor onto, and no
inverse function exists.
Probability Space
• Essential to developing a conceptual understanding of
problems involving probability.
• It is a mathematical construct consisting of three
components:
a) Sample space S
b) Sigma algebra of events A
c) Probability measure P
Random Experiment
• A well-defined experiment in which the experimental outcome
cannot be predicted before conducting the experiment.

Sample Space (S)

• A set consisting of the possible outcomes for a random
experiment.

Event (E)
• A subset of the sample space from a random experiment.
Sigma algebra of events (A)
• A set of events from a random experiment that satisfies the following
properties:

1) If E  A, then EC  A
2) If E1  A and E2  A, then E1  E2  A
3) If E1  A and E2  A, then E1  E2  A
4) SA

• In general, if the discrete sample space S contains k outcomes,

the associated sigma algebra of events A will contain 2k events.
Probability Measure (P)
• A function having domain A, codomain [0, 1]  R, and rule of
correspondence P such that the following properties hold:

1) P(E)  0 for all E  A

2) P(S) = 1
3) P(E1  E2) = P(E1) + P(E2) when E1  E2 = { }
Example:
A discrete random experiment consisting of rolling a three-sided die.
The sides of the die are labeled as B, C and D. When the die is rolled, it
is found out that side B shows one-sixth of the time, C shows one-third
of the time and side D shows one-half of the time. It is desired to
construct a probability space for this random experiment.

Solution: Possible outcomes: Side B showing; side C showing; or side

D showing.

Sample space S = { B, C, D }
Sigma algebra of events A:

Since S contains 3 elements, then A must contain 23 events which is

equal to 8 including the null set and S.
Therefore, A = { { }, {B}, {C}, {D}, {B, C}, {B, D}, {C, D}, {B, C, D} }

Probability measure P:
Random Variable
• A means of associating real numbers with experimental outcomes.

Let  = (S, A, P) NOTE:

where:  – probability space
X has:
S – sample space • S – domain set
A – sigma algebra of events • R – codomain set
P – probability measure • X:S→R – rule of correspondence such that
{ X  x } = {s | s  S and X(s)  x } is an event
in A for all x in R.
Let X = random variable, which is simply a To qualify as a random variable, the set of
function that assigns to each experimental outcomes s that satisfy X(s)  x must be an
outcome s  S a real number X(s). event in A for all values of x  R.
Functions associated with random variable X

1. Cumulative probability distribution function F(x)

• Also referred to as Cumulative Distribution Function or Distribution Function.
• It has:
➢ Domain R
➢ Codomain [0, 1]  R
➢ Rule of correspondence F:R→[0, 1] such that F(x) = P({X(s)  x})
Properties of F(x):
Example: Three-sided die
where u(x) is the unit
step function defined
as:
Functions associated with random variable X

2. Probability density function f(x)

• Also referred to as density function.
• It has:
➢ Domain R
➢ Codomain R
➢ Rule of correspondence p:R→R such that f(x) = dF(x) / dx.
Properties of f(x):

1) f(x)  0 for all x  R.

∞
2) The total area under the f(x) curve is 1. That is ‫׬‬−∞ 𝑓 𝑥 𝑑𝑥 = 1.
𝑥
3) The distribution function is related to the density function via integration; that is, 𝐹 𝑥 = ‫׬‬−∞ 𝑓 𝜎 𝑑𝜎.
𝑥
4) P({x1< X  x2} ) = ‫ 𝑥׬‬2 𝑓 𝑥 𝑑𝑥.
1
Theoretical Distribution Functions Used Commonly in M&S

1. Uniform Distribution
• Often used when a random variable having equally likely values over a finite
range is needed.
• Denoted as UNIFORM (a, b)
where: a – minimum value of the random variable
b – maximum value of the random variable
a, b  R and - < a < b < 
• The probability density function for this distribution is

f(x)
Theoretical Distribution Functions Used Commonly in M&S

2. Triangular Distribution
• Used in situations in which the exact form of the distribution is not known, but approximate
values are known for the minimum, maximum and most likely values.
• Denoted as TRIANGULAR(a, m, b)
where: a – minimum value of the distribution f(x)
b – maximum value of distribution
m – mode or most likely value of the distribution
a, m, b  R and - < a < m < b < 
• The probability density function for this distribution is:

f(x)
Theoretical Distribution Functions Used Commonly in M&S

3. Exponential Distribution
• Used to model interarrival times for entities arriving at a system.
• Denoted as EXPONENTIAL(m)
where: m – the mean of the distribution
m  R and 0 < m <  − 𝑥
𝑒 ,𝑥 ≥ 0
• The probability density function for this distribution is: 𝑓 𝑥 = ൝
0 , 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

f(x)  - average number of events in

m one unit of time
1 𝑃 𝑋 > 𝑥 = 𝑒 −𝑥
𝑚=
0 x
 𝑃 𝑋 < 𝑥 = 1 − 𝑒 −𝑥
units of time
Theoretical Distribution Functions Used Commonly in M&S

4. Normal Distribution
• Used to represent variability of quantities about some average value.
• This distribution also represents the distribution of a random variable formed as the sum of a
large number of other random variables, even when the other random variables are not
normally distributed.
• Denoted as NORMAL(m, )
where: m – the mean of the distribution
 - standard deviation of the distribution
m,   R, - < m < , and 0 <  < . 1 1 𝑥−𝜇 2
• The probability density function for this distribution is 𝑓 𝑥 = 𝑒 2 𝜎 )
− (
𝜎 2𝜋
f(x)
Generation of Random Variates

Random Variates
• The independent samples selected from a random variable having a theoretical
distribution.

Note: The capability to quickly and efficiently generate random variates is an

essential requirement for all DES software systems.

Steps in generating random variates:

1. A random number is generated.
2. The random number is transformed to a random variate.
Generation of Random Variates

Random Variates
• The independent samples selected from a random variable having a theoretical distribution.

Note: The capability to quickly and efficiently generate random variates is an essential requirement
for all DES software systems.

Steps in generating random variates:

1. A random number is generated.
• A random number is defined as a sample selected randomly from the uniform distribution
UNIFORM(0, 1).
• A random number is a real number between zero and one, wherein at each random
number draw, each real number in the range [0, 1] has an equal chance of being selected.

2. The random number is transformed to a random variate.

Generation of Random Variates
• In simulation, random numbers are generated on a computer using a recursive algorithm. Most
common algorithms used to simulate random number generation are called Linear Congruential
Generators (LCG).
• Using an LCG, a sequence of integers is generated using the recursive formula:

constant multiplier increment modulus

Note: a, c , m → nonnegative integers

Z0 → seed
Z 1 , Z2, Z3, … → sequence of integers generated, called linear congruential sequences,
and limited to the range 0  Zk  m – 1. (Pseudorandom numbers)

The output of the LCG, Uk, is obtained by dividing each integer Zk by m.

Note:
Each seed Z0 will produce a sequence of p distinct integers, p  m, and then the sequence will begin
to repeat. Generators with p = m are called full-cycle or full-period generators.

Thus, to be useful in simulation, the constants in the LCG and the initial seed must be selected to
produce a very long cycle length.

Source: Olivar, A. (2015). Stochastic Simulation of Random Numbers Using Linear Congruential Method. Unpublished Thesis, AMA University
Source: Olivar, A. (2015). Stochastic Simulation of Random Numbers Using Linear Congruential Method. Unpublished Thesis, AMA University
Components in the development of a system model

1. Structural Modeling
• The entities, locations, resources and processes describing the structure and operation of
the system are defined.
• The result is a mathematical or logical representation (i.e., Queuing System Model)
2. Data Modeling
• Descriptive data required to execute the structural model are developed.
• Data include entity interarrival times, process service times, resource schedules and
failure rates, entity travel times and all of the other system data needed to describe the
operation of the system quantitatively.
• The data model is used to generate the many random variate streams that must be
supplied during the simulation of the system model.
Choice for generating the random variate stream required for
simulation

1. Direct Use of Sample Data

2. Use of Empirical Distribution
3. Use of Theoretical Distribution
Direct Use of Sample Data
• Sample values are stored in a data file.

Advantage: Generates only legal and realistic random variates.

Disadvantages:
a. In a limited set of sample values, legal and realistic values that occur
infrequently may not be present.
b. Making repeated calls to data memory to retrieve stored sample values is
computationally very slow.
c. Each simulation execution may require a large number of random variate
draws.

Note: This approach is not often used due to practical limitations.

Use of Empirical Distribution
• A nonparametric distribution constructed to exactly represent the available
set of sample values.
Advantages:
a. Facilitates the generation of random variate stream of unlimited length.
b. It can be used to transform a stream of random numbers of arbitrary length
to the desired random variate stream.
c. Each random variate will be a legal and realistic value.

Disadvantage: It cannot generate outlying values that may actually occur but
are only infrequently observed.

Note: Generally used when a good fit to the sample data using a theoretical
distribution cannot be found.
Use of Theoretical Distribution
• A distribution function that is defined mathematically in terms of distribution
parameters.
• Parameters are often related to the position, scale or shape of the
theoretical distribution.

Examples:
a. Exponential Distribution
b. Normal Distribution
c. Triangular Distribution
Use of Theoretical Distribution
Advantages:
a. Smooths irregularities resulting from using a limited number of sample
values.
b. Use of this approach is computational efficient.

Disadvantage: The computational effort required to fit a distribution to sample

data and, for some sample data sets, the difficulty in finding a good fit.
Output Data Analysis
• The process of analyzing simulation output data to produce useful information concerning
system performance.

Steps:
a. A simple queuing system example is presented and used to illustrate why output data
analysis is necessary.
b. The confidence interval estimate of the mean is described.
c. Some of the more complex output data analysis situations are identified , and the
approaches for addressing these situations are briefly discussed.
Formative Assessment 2
Check Canvas for instructions
 SUBTOPIC 2
DISCRETE EVENT SIMULATION
• Understand concepts on discrete event simulation.
• Discuss the construction and performance of a queuing
system model.
• Illustrate a simulation methodology by describing a
general DES framework and further illustrate the
mechanics of the simulation.
After the completion of the topics and laboratory activities, students can:

• Understand concepts on discrete event simulation.

• Discuss the construction and performance of a queuing system model.
• Illustrate a simulation methodology by describing a general DES
framework and further illustrate the mechanics of the simulation.
What is Discrete Event Simulation (DES) ?
• Modeling of a system as it evolves over time by a representation
where the state variables change instantaneously at separated
points in time. (Law:Simulation Modeling and Analysis)
• It is a model of a physical system that has changes at precise points
in simulated time.
• It is a method of simulating the behavior and performance of a real-
life process, facility or system.

NOTE: DES is stochastic, dynamic and discrete.

Example:

TIME DELAY or INTERVAL

This will model the actual

movement of the train

In Discrete Event Modeling, the movement of the train from Point A to Point B will be modelled by these two events:
Other Examples:
Other Examples:
 In DES, we calculate the
solution for every change of
events. For example, in a
queuing system, we
calculate the system when
an entity arrives, or when
an entity leaves the system.
DES finds its applications
in simulating queuing of
people in a hospital,
queuing of cars in an
intersection, queuing of
products in manufacturing
line, etc.
1. Entity
• An object that can interact with
each other or competes with
each other for a resource.

Resource
Entity Entity
2. Attribute
• Variable (s) that pertains to an individual entity and is carried along with the
entity as it goes through the system.

NOTE:

In DES, there is some sort of a memory that records

experience of an entity as it goes through a system.
This memory is represented by changes in attributes.
3. Resources
Capacity of a Resource:
Number of entities to be served at a time.

Example: ATM capacity is 1

Bank with three tellers: Bank capacity is 3

Resource Utilization
- one of the outcomes of DES

Busy time / Total time

4. Queue
• The list that an entity enters if a resource is occupied.
• Queue logic: FIFO / LIFO / HVF

Resource
5. Events
• Any change in the state of the system.
Examples:
Entity is created
Entity interacts/competes with another entity
Entity leaves the system
Discrete event simulation is generally carried out by a software designed in high level
programming languages such as Pascal, C++, or any specialized simulation
language.
Following are the five key features:
❑ Entities − These are the representation of real elements like the parts of machines.
❑ Relationships − It means to link entities together.
❑ Simulation Executive − It is responsible for controlling the advance time and
executing discrete events.
❑ Random Number Generator − It helps to simulate different data coming into the
simulation model.
❑ Results & Statistics − It validates the model and provides its performance
measures.
Time Graph Representation

Every system depends on a time parameter. In a graphical representation it is

referred to as clock time or time counter and initially it is set to zero.

Time is updated based on the following two factors:

❑ Time Slicing − It is the time defined by a model for each event until the absence
of any event.

❑ Next Event − It is the event defined by the model for the next event to be
executed instead of a time interval. It is more efficient than Time Slicing.
Common Terminology Associated with Discrete Event Simulation

• System State. A characterization of the state of the system at a particular

moment; expressed as state variables.
• Simulation Clock. A tool that provides the elapsed real -world time.
• Next-Event List. An ordered list of times that each event will take place.
• Statistical Counter or Accumulator. A tool that both records and expresses a
system’s evolving performance; it accumulates information as the simulation
unfolds.
• Initialization Subprogram. A protocol utilized in the initialization of the
simulation, usually setting the start time to zero.
• Timing Subprogram. A protocol that, drawing from a next-event list, sets the next
event and progresses the simulation clock to the moment when an event is to
happen.
Common Terminology Associated with Discrete Event Simulation
• Event Subprogram. A protocol that launches a routine that updates the state
of the system with the occurrence of each event.

• Library Subprogram. A protocol used to produce random observations drawn

generally from predetermined probability distributions.

• Report Generator. A tool that calculates and reports statistics that describe the
performance of the system.

• Main Program. A routine that coordinates the concert of subordinate routines,

executing these in the correct sequence. It initializes the timing subprogram
that determines the subsequent event, passes control to the related event
subprogram, and updates the system state. This routine verifies for termination
and triggers the report generator once the simulation ends.
DES Simulation Flowchart
Classification of Systems

1. Discrete Systems
• The state variables are updated instantaneously at specified times
(e.g., t1 , t2 , t3 … ).

Example:
A lunch queue and server utilization at a local fast - food restaurant
throughout the peak midday business hours.

2. Continuous Systems
• The state variables change continuously with the progression of time.

Example: State variable “ volume ” exhibits continuous change over time as

a tank is filled with fluid.
Approaches for Advancing the Simulation Clock

1. Fixed-increment Advance
• Uncommon in simulation software.
• Initializes at time zero and then advances at fixed time increments.
• Events such as an entity’s arrival to, or departure from, the system, as
generated by the probability distribution and recorded in the event list, may
fall between successive moments that demarcate the time interval.
• The updating of the state variables takes place not with the occurrence of an
event, but with the passing of a time increment moment.
• For the purpose of recording data and updating state variables, the event is
artificially forced in sync with the interval.
• The size of the time increment is subjective. The choice of increment has
implications for the performance measures of the system.
Approaches for Advancing the Simulation Clock

2. Next-Event Time Advance

• Employed within most simulation software.
• Initializes the simulation clock at zero, progresses to the most proximate,
forthcoming event as tendered by the event list, and then updates the state
variables.
• Once this is complete, the simulation clock is again progressed to the
moment of the next most forthcoming event, and the state of the system is
again revised.
• The updating of the state variables is in sync with the occurrence of events.
• Omits periods of inactivity.
 - Represents a certain number of customers waiting for service.

Examples: Banks Ships

Post Offices Railway Ticket Counters

What causes queues?

* Arrival Variation * Service Variation

Transaction very brief

Transaction took more time

Transaction simultaneous
with other customer
 • A representation that captures and quantifies the
phenomenon of waiting in lines.

Basic Elements
1. Entities – Customers or objects
2. Servers – Persons or production stations that treat or interact
with the entity
3. Queues – Holding or waiting position of entities
1. Arrivals of Customers
• It defines the way customers enter into the system.
• Mostly the arrivals are random with random intervals between two adjacent
arrivals.
• The arrivals are described by a random distribution of intervals referred to as Arrival
Pattern.
2. Service Mechanism
• A customer who arrives and finds the server idle enters service immediately, and
the service time of the successive customers are independent and identically
distributed (IID) random variables that are independent of inter arrival times.
• A customer who arrives and finds the server busy joins the end of a single queue.

NOTE: IID means that the arrival times of the entities are independent, or
autonomous, of each other and that the probability distribution is identical, or alike,
for all entities.
3. Queuing Discipline
• It represents the way the queue is organized or it is the order in
which service is provided such as:
o FIFO – First in, First Out → also called FCFS
o FIFO – Last in, First Out → also called LCFS.
o Priority Queue → It maybe viewed as a number of queues for
various priorities
o SIRO → Serve in random order
• The order that either customers or objects are selected from a
queue to advance to receive service.

Two common disciplines:

1. FCFS – First come, first served
2. STP - Shortest process time
1. Analytic Model
- using differential equations
- for simple cases

Example: Model for a fast food restaurant

2. Simulation Model
- used for complex systems with multiple stations

Example: Low Observable (LO) aircraft maintenance system

1. Balance two costs
• Service cost → Cost of providing service
• Waiting cost → Cost of not providing service
* Bulk : Leave without buying because the line is too long
* Buy and never return
2. Goal is to minimize cost
• Jockeying
✓ When the customer enters one line and then switches to a
different one in an effort to reduce the waiting time
• Reneging
✓ The customer enters the line but decides to leave before being
served
• Balking
✓ The customer decides not to enter the waiting line
• Faffing
✓ Time delay when a person gathers his things after paying at the
checkout
• FIFO
Arrival and Service Time
Stochastic
• Random
• Non deterministic
Arrival Process
• Characterized by the interarrival time, or the interval of time between
successive customers or objects entering the system in preparation to receive
service

Arrival time
• Independent, or autonomous, of each other and that the probability
distribution is identical for all entities
1. Arrival rate → How fast are customers will show up ?

2. Service rate → How fast will the server handle the customer ?
3. Number of servers → How many servers are needed to provide
service to the customers ?
4. Maximum queue length → Optional
5. Population size → Optional
1. Single Server, Single Phase → (McDonald’s, ATMs)
- Providing service to a single line
2. Single Server, Multi Phase → (Chipotle)
3. Multiple Server, Single Phase → (Chick-fil-A)
- Multiple servers serving a single line
4. Multiple Server, Multiple Phase → (Laundromat)
- Multiple servers serving a single line
Formulas to be used
1. Time since last arrival for (i + 1) customer = Arrival time of ith customer – Arrival time of (i – 1)th customer

2. Time customer wait in queue for ith customer = Arrival time of ith customer – Time service begin of ith customer

3. Time service ends of ith customer = Time service begin of ith customer (min) + Service Time of ith customer

4. Time customer spends in system = Time service ends of ith customer – Arrival time of ith customer

5. Idle time of server for ith customer = Time service begins of ith customer – Time service ends of (i – 1)th
customer

6. Time service begins:

For first customer, it is zero (always).

For rest of customer, it is a maximum between arrival of ith customer and time service ends
of (i – 1)th customer. i.e. Max. (Arrival of (ith) customer, Time service ends of (i – 1)th
customer)
Example:

Calculate the different parameter for the single server queueing

system when inter-arrival time and service times are given below:
Customer 1 2 3 4 5 6
Inter-arrival time (IAT) OR (AT) - 4 1 4 2 4
Arrival time on clock 0 4 5 9 11 15
Service Time 2 3 1 4 2 3
Solution:
Customer Time Arrival Service Time Waiting Time Time Idle
service Time time service in service customer time
last begin queue ends spends of
arrival in queue
system
1 - 0 2 0 (0 – 0) = 0 (0 + 2) = 2 (2 – 0) = 2 0–0=0
2 4 4 3 4 (4 – 4) = 0 (4 + 3) = 7 (7 – 4) = 3 4–2=2
3 1 5 1 7 (7 – 5) = 2 (7 + 1) = 8 (8 – 7) + 2 = 3 7–7=0
or
8–5=3
4 4 9 4 9 (9 – 9) = 0 (9 + 4) = 13 (13 – 9) = 4 9–8=1
5 2 11 2 13 (13 – 11) = 2 (13 + 2) = 15 (15 – 13) + 2 = 4 13 – 13 = 0
or
15 – 11 = 4
6 4 15 3 15 (15 – 15) = 0 (15 + 3) = 18 (18 – 15) = 3 15 – 15 = 0
15 15 4 19 3
Formative Assessment 3
Check Canvas for instructions
Summative Assessment 2
Check Canvas for instructions
Online:
www. tutorialspoint.com

E-Books and Textbooks:

• Sokolowski, John A. & Banks, Catherine M. (2010). Modeling and Simulation
Fundamentals Theoretical Underpinnings and Practical Domains. John Wiley &
Sons