DISCRETE-EVENT SIMULATION Entity: Any object or component in the system that

requires explicit representation in the model (e.g., a

2.1 DISCRETE-EVENT SIMULATION server, a customer, a machine).
Discrete-event simulation is the modeling of Attributes: The properties of a given entity (e.g., the
systems in which the state variable changes only at a priority of a waiting customer, the routing of a job
discrete set of point in time. The simulation models through job shop).
are analyzed by numerical rather than analytical
List: A collection of (permanently or temporarily)
methods.
associated entities, ordered in some logical fashion.
• Analytical methods employ the deductive (such as calling customers currently in a waiting line)
reasoning of mathematics to “solve” the model.
Event: An instantaneous occurrence that changes the
• Numerical methods employ computational
state of a system (such as an arrival of a new
procedures to “solve” mathematical models.
customer)
In case of simulation models, which employ
Event notice: A record of an event to occur at the
computational procedures to “solve” mathematical
current or some future time, along with any associated
methods, models are “run” rather than solved-that is,
data necessary to execute the event; at a minimum,
an artificial history of the system is generated from the
the record includes the event type and the event time.
model assumptions, and observations are collected
to be analyzed and to estimate the true system Event list: A list of event notices for future events,
performance measures. ordered by time of occurrence; also known as future
event list (FEL)
Real-world simulation models are rather large,
and the amount of data stored and manipulated is Activity: A duration of time specified length (e.g., a
vast, so such runs are usually conducted with aid of a service time or interarrival time), which is known when
computer. it begins
The concept of a system and a model of a Delay: A duration of time of unspecified indefinite
system discussed in the previous chapter. This length, which is not known until it ends (e.g., a
chapter deals exclusively with dynamic, stochastic customer’s delay in a last-in-first-out waiting line
systems (i.e., involving time and containing random which, when it begins, depends on future arrivals)
elements) that change in a discrete manner. This
Clock: A variable representing simulated time, called
section expands on these concepts and proposes a
clock in the example to follow.
framework for the development of a discrete-event
model of a system. The major concept are briefly Different simulation packages use different
defined and then illustrated by examples: terminology for the same or similar concepts – for
example, lists are sometimes called sets, queues, or
System: A collection of entities (e.g., people and
chains. Sets or lists are used to hold both entities and
machine) that interact together over time to
event notices.
accomplish one or more goals.
A discrete-event simulation is the modeling
Model: An abstract representation of a system,
over time of a system all of whose state changes occur
usually containing structural, logical, or mathematical
at discrete point in time – those points when an event
relationship that describes a system in terms of state,
occurs. A discrete event simulation proceeds by
entities and their attributes, sets, processes, events,
producing a sequence of system snapshots that
activities and delays.
represent the evolution of the system through time.
System rate: A collection of variables that contain all
the information necessary to describe the system at
any time.
 2.2 THE EVENT SCHEDULING/TIME ADVANCE
ALGORITHM

The mechanism for advancing simulation time

and guaranteeing that all event occur in correct
chronological order is based on the future event list
(FEL). This list contains all event notices for events
that have been scheduled to occur at a future time.
Scheduling a future event mean that, at the instant an
activity begins, its duration is computed or drawn as a
sample from a statistical distribution; and that the
end-activity event, together with its event time, is
placed on the future-event list. In the real world, most
future events are not scheduled but merely happen-
such as random breakdowns or random arrivals. In the
model, such random events are represented by the Some sequence of the uniformity and
end of some activity, which in turn is represented by a independence properties are the following:
statistical distribution.
1. If the interval [0, 1] is divided into n classes, or
At any given time t, the FEL contains all subinterval is N/n, where N is the total number
previously scheduled future events and their of observations.
associated event times (called t1, t2 … ). The FEL is 2. The probability of observing a value in a
ordered by event time, meaning that the events are particular interval is independent of the
arranged chronologically-that is event times satisfy. previous values drawn.

In the next section we describe desirable

properties of methods that produce random numbers
to drive simulations.

3.1 RANDOM-NUMBER GENERATION 3.2 GENERATION OF PSEUDO-RANDOM NUMBERS

Random numbers are a necessary basic Notice that the title of this section has the
ingredient in the simulation of almost all discrete word “pseudo” in it. “Pseudo” means false, so false
systems. Most computer languages have a random numbers are being generated! In this
subroutine, object, or function that will generate a instance, “pseudo” is used to imply that the very act of
random number. generating random numbers by a known method
removes the potential for true randomness. If the
Random numbers are used to generate event method is known, the set of random numbers can be
times and other random variables in simulation repeated. Then an argument can be made that the
languages. In this chapter, the generation of random numbers are not truly random. The goal of any
numbers and their subsequent testing for generation scheme, however is to produce a
randomness is described. sequence of numbers between 0 and 1 that simulates
3.2 PROPERTIES OF RANDOM NUMBER or imitates, the ideal properties of uniform distribution
and independence as closely as possible.
A sequence of random number R1, R2 ….,
To be sure, in the generation of pseudo-
must have two important statistical properties:
random numbers, certain problems or errors can
uniformity and independence.
occur. These errors, or departures from ideal
Each random number Ri must be an randomness, all related to the properties stated
independent sample drawn from a continuous previously.
uniform distribution between zero and 1.
 Some example of such departures include the should be possible to easily specify different
following: starting points, widely separated, within the
sequence.
1. The generated numbers might not be
5. Most important, the generated random numbers
uniformly distributed.
should closely approximate the ideal statistical
2. The generated numbers might be discrete-
properties of uniformity and independence.
valued instead of continuous-valued.
3. The mean of generated numbers might be too Inventing techniques that seem to generate
high or too low. random numbers is easy; inventing techniques that
4. The variance of the generated numbers might really produce sequences that appear to be
be too high or too low. independent, uniformly distributed random numbers
5. There might be dependence as measured, for is incredibly difficult. There is now a vast literature and
instance, by autocorrelation. rich theory on the topic, and many hours of testing
have been devoted to establishing the properties of
Departures from uniformity and
various generators.
independence for a particular generation scheme
often can be detected by tests. If departures are Even when a technique is known to be
detected, the generation scheme should be dropped theoretically sound, it is seldom easy to implement it
in favor of an acceptable generator. in a way that will be fast and portable. The goal of this
chapter is to make the reader aware of the central
Usually random numbers are generated by a
issues in random-number generation, to enhance
digital computer as part of the simulation. There are
understanding, and to show some of the techniques
numerous methods that can be used to generate the
that are used by those working in this area.
values. Before we describe some of these methods,
there are several important considerations that we Example of a generated Pseudo-Random Number
should mention:

1. The method should be fast. Individual

computations are inexpensive, but simulation
could require many millions of random numbers.
The total cost can be managed by selecting a
computationally efficient method of random-
number generation.
2. The method should be portable to different
computers-and, ideally, to different
programming languages. This is desirable so that
simulation program will produce the same
results wherever it is executed.
3. The method should have a sufficiently long cycle.
The cycle length, or period, represents the length
of the random number sequence before previous
number begin to repeat themselves in an earlier
order. Thus, if 10,000 events are to be generated,
the period should be many times that long.
4. The random numbers should be repeatable.
Given the starting point (or conditions) it should
be possible to generate the same set of random
numbers, completely independent of the system
that is being simulated. This is helpful for
debugging purposes and facilitates comparisons
between systems. For the same reasons, it
3.3 TEST FOR RANDOM NUMBERS 3.3.2 TEST FOR AUTOCORRELATION

The desirable properties of random numbers- Test for Autocorrelation

uniformity and independence- were discussed
The test for autocorrelation are concerned
previously. To check on whether these desirable
with the dependence between number in a sequence.
properties have been achieved, a number of tests can
As an example, consider the following sequence of
be performed. (Fortunately, the appropriate tests have
numbers, read from left to right.
already been conducted for most commercial
simulation software). The tests can be placed in two
categories, according to the properties of interest:
uniformity and independence. A brief description of
two types of tests is given in this section:
From a visual inspection, these numbers
1. Frequency test. Use of Kolmogorov-Smirnov or appear random, and they would probably pass all the
the chi-square test to compare the distribution of tests presented to this point.
the set of numbers generated to a uniform
However, an examination of the 5th , 10th ,
distribution.
15th, (Every five numbers beginning with the fifth), and
2. Autocorrelation test. Tests the correlation
so on, indicates a very large number in that position.
between numbers and compares the sample
correlation to the desired correlation, zero. Now, 30 numbers is rather a small sample size
on which to reject a random number generator, but the
3.3.1 FREQUENCY TEST
notion is that the numbers in the sequence might be
Frequency test related.

A basic test that should always be performed In this section, a method for discovering
to validate a new generator is the test of uniformity. whether such a relationship exists is described. The
Two different methods of testing are the Kolmogorov- relationship would not have to be all high numbers. It
Smirmov and the chi-square test. Both test measures is possible to have all low numbers in the locations
the degree of agreement between the distribution of a being examined, or the numbers could alternate from
sample of generated random numbers and the very low to very high.
theoretical uniform distribution. Both tests are based 3.4 MECHANICS OF DISCRETE SIMULATION
on the null hypothesis of no significant difference
between the sample distribution and the theoretical As an example, let’s think about a logistics chain
distribution. involving a truck:

1. it arrives at a warehouse;
2. it enters a loading gate;
3. it unloads its supplies;
4. and then it leaves the warehouse. In the
example above, you will see that each step is a
separate — i.e., discrete — event. In order to
simulate this process, you will typically use a
discrete event simulation model.
 MANUFACTURING AND MATERIAL HANDLING 7. Production schedules
SIMULATIONS
a) Made-to-stock
Manufacturing and Material Handling Simulations b) Made-to-order
Customer order
As do all modeling projects, manufacturing Line items and quantities
and material handling simulation projects need to
8. Production control
address the issue of scope and level of detail.
Consider scope as analogous to breadth and level of a. Assignment of jobs to work areas
detail as analogous to depth. b. Task selection at workcenters
c. Routing decisions
The proper scope and level of details should
be determined by the objective of the study and the 9. Supplies
questions being asked.
a) Ordering
On the other hand, level of detail could be b) Receipt and storage
restrained by the availability of input data and the c) Delivery to workcenters
knowledge of how system components work.
10. Storage
MODELS OF MANUFACTURING SYSTEM
a. Supplies
Models of manufacturing systems might have b. Spare parts
to take into account a number of characteristics of c. Work-in-process (WIP)
such systems, including some of the following: d. Finished goods

1. Physical layout 11. Packing and shipping

2. Labor a. Order consolidation

b. Paperwork
a. Shift schedule c. Loading of trailers
b. Job duties and certification
MODELS OF MATERIAL HANDLING SYSTEMS
3. Equipment
In manufacturing systems, it is not unusual for
a. Rates and capacities
80 to 50% of an item’s total time in system to be
b. Breakdowns
expended on material handling or on waiting for
Time to failure
material handling to occur.
Time to repair
Resources needed for repair This Work-in-progress represents a vast
investment, and reductions in WIP and associated
4. Maintenance
delays can result in large cost savings. Therefore, for
a. Preventive maintenance (PM) schedule some studies, detailed material-handling simulations
b. Time and resources required are cost effective.
c. Tooling and fixtures
Storage systems
5. Work centers
a) Pallet storage
a) Processing b) Case storage
b) Assembly c) Small-part storage (totes)
c) Disassembly d) Oversize items
e) Rack storage or block stacked
6. Product
f) Automated storage and retrieval systems
a. Product flow, routing and resources needed (AS/RS) with storage retrieval machines (SRM)
b. Bill of materials
GOALS AND PERFORMANCE MEASURES Modeling downtime and failures

The purpose of simulation is insight, not Unscheduled random downtimes can have a
numbers. major effect on the performance of manufacturing
systems. Many authors have discussed the proper
Those who purchase and use simulation
modeling of downtime data.
software and services want to gain insight and
understanding into how a new or modified system will There are a number of alternatives for modeling
work. random unscheduled downtime, some better than the
others:
Will it meet throughput expectations? What
happens to response time at peak periods? 1. Ignore it
2. Do not model it explicitly, but increase
Is the system resilient to short-term surges?
processing times in appropriate proportion
What is the recovery time when short-term surges
3. Use constant values for the time to failure and
causes congestion and queueing?
time to repair
Simulation are expected to provide numeric 4. Use statistical distribution for time to failure
measures of performance, such as throughput under and time to repair
a given set of conditions, but the major benefit of
Alternative one (1) is not the suggested general
simulation comes from the insight and understanding
approach. This is certainly an irresponsible modeling
gained regarding system operations.
technique if downtimes have an impact on the results,
The major goals of manufacturing-simulation as they do in almost all situations. One situation in
model are to identify problem areas and quantify which ignoring downtimes could be appropriate, with
system performance. Common measures of system full knowledge of the customer, is to leave out the
performance include the following: catastrophic downtimes that occur rarely and leave a
production plant or line for a long period of time.
• Throughput under average and peak loads;
• System Cycle time (how long it takes to In other words, the model would incorporate
produce one part); normal downtimes but ignore those catastrophic
• Utilization of resources, labor and machines; downtimes, such as general power failures, snow
• Bottlenecks choke points; storms, cyclones and hurricanes, that occur rarely,
• Queueing at work locations; but stop all production when they occur.
• Queueing and delays caused by material- The second possibility, to factor into the model
handling devices and systems; the effect of downtimes by adjusting processing times
• WIP storage needs; applied to each job or part, might be an acceptable
• Staffing requirement; approximation under limited circumstances. If each
• Effectiveness of scheduling systems; job or part is subjected to a large number of small
• Effectiveness of control systems; delays associated with downtime of equipment or
tools, then the total of such delays may be added to
Issues in manufacturing and material-handling
the pure processing time to arrive at an adjusted
simulations
processing time.
There are a number of modeling issues If total delay time and pure processing time
especially important for the achievement of accurate are random in nature, then an appropriate statistical
and valid simulation models of manufacturing and distribution should be used for the total adjusted
material-handling systems. Two of these issues are processing time. If the pure processing time is
the proper modeling of downtimes and whether, for constant while the total delay time in one cycle is
some inputs, to use actual system data or a statistical random and variable, it is almost never accurate to
model of those inputs. adjust the processing time by a constant factor.
 The third possibility, using constant durations
for time to failure and time to repair, might be
appropriate when, for example the downtime is
actually due to preventive maintenance that is on a
fixed schedule.

In almost all other circumstances, the fourth

possibility, modeling time to failure and time to repair
by appropriate statistical distribution when data is
lacking, a reasonable assumption based on the
physical nature of the causes of downtimes.

Time to failure

Time to failure can be measured in a number of

different ways

1. By wall-clock time;
2. By machine or equipment busy time;
3. By number of cycle times;
4. By number of items produced.

Breakdown/Failure

Breakdowns or failures can be based on

clock time, actual usage or cycles. Note that the word
breakdown or failure is used, even though preventive
maintenance could be the reason for a downtime. As
mentioned, breakdowns or failures can be
probabilistic or deterministic in duration.

Actual usage breakdown are based on time during

which the resources is used.

Clock-time breakdowns might be associated with

scheduled maintenance.

Cycle breakdown or failures are based on the

number of times the resources is used.

Another issue is what happens to a part at a

machine when the breakdown or failure occurs.
Possibilities include scrapping the part, rework, or
simply continuing processing after repair. In some
cases-for example, when the preventive maintenance
is due- the part in the machine may complete
processing before the repair (or maintenance activity)
begins.

1. As a pure time delay (no resources required);

2. As a wait time for a resource (e.g.,
maintenance person) plus a time delay for
actual repair.