This article was downloaded by: [Univ Autonoma De Yucatan]

On: 14 March 2014, At: 18:03

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office:
Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Construction Management and Economics

Publication details, including instructions for authors and subscription
information:
http://www.tandfonline.com/loi/rcme20

Three methods for verifying and validating the

simulation of a construction operation
Jonathan Jingsheng Shi
Published online: 21 Oct 2010.

To cite this article: Jonathan Jingsheng Shi (2002) Three methods for verifying and validating the
simulation of a construction operation, Construction Management and Economics, 20:6, 483-491, DOI:
10.1080/01446190210151032
To link to this article: http://dx.doi.org/10.1080/01446190210151032

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content)
contained in the publications on our platform. However, Taylor & Francis, our agents, and our
licensors make no representations or warranties whatsoever as to the accuracy, completeness, or
suitability for any purpose of the Content. Any opinions and views expressed in this publication are
the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis.
The accuracy of the Content should not be relied upon and should be independently verified with
primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever
caused arising directly or indirectly in connection with, in relation to or arising out of the use of the
Content.
This article may be used for research, teaching, and private study purposes. Any substantial
or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or
distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can
be found at http://www.tandfonline.com/page/terms-and-conditions

Construction Management and Economics (2002) 20, 483491

Three methods for verifying and validating

the simulation of a construction operation
JONATHAN JINGSHENG SHI*
Department of Civil and Architectural Engineering, Illinois Institute of Technology, 3201 S. Dearborn Street,
Chicago, IL 606163793, USA

Downloaded by [Univ Autonoma De Yucatan] at 18:03 14 March 2014

Received 30 May 2001; accepted 19 April 2002

A simulation model must be veri ed to con rm that it describes correctly its intended real world process
under study; moreover, the simulation results obtained must be a valid representation of the process. This
study presents three activity-based white-box methods for assisting a user in verifying and validating construction simulations. The rst method reports a simulation by listing all activities in the chronological order
of their executions, so that a user can contrast the simulated progress against the actual progress in the real
world. The second method summarizes the operating counts and mean durations of all activities over the
simulated time period, to enable a user to evaluate whether all activities have been executed correctly during
simulation. The third method generates an activity cycle report for any selected resource entity, so that a
user can examine whether the entity is moving in the correct logical and chronological order during simulation. The three methods can be used jointly to debug a simulation model, so as to con rm that the simulation is correctly conducted and the obtained results are valid. The three methods are implemented in a
simulation system in the form of corresponding reports. A concreting process is employed to illustrate these
methods.
Keywords: Computer simulation, simulation modelling, validation, veri cation, construction planning, simulation of construction operations, construction process

Background
Computer simulation has been growing rapidly with
the advancement of computer technology and has
become one of the most popular techniques for practitioners in operations research and in the manufacturing industry (Pidd, 1988; Paul, 1991). The rst
application of simulation in construction was reported
by Teicholz (1963). Since the development of Cyclone
(Halpin, 1977), extensive research efforts have resulted
in many construction simulation systems such as
Insight (Paulson, 1978), Resque (Chang, 1987), UMCyclone (Ioannou, 1989), Coops (Liu and Ioannou,
1992), Disco (Huang et al., 1994), Cipros (Tommelein
and Odeh, 1994), Stroboscope (Martinez and
Ioannou, 1999), Simphony (Hajjar and AbouRizk,
*Author for correspondence. e-mail: jonathan.shi @ iit.edu

1999), and ABC (Shi, 1999). Actual applications

reported signi cant improvement in construction productivity. For example, an international contractor
(Dragados) has used Cyclone for over 30 projects and
has recorded productivity improvements ranging from
30% to 300%. For every hour of analyst time used, a
saving of $2000 was realized (Halpin and Martinez,
1999).
Simulation is the technology of experimenting on a
computer with a mathematical and/or logical model
built for a real world system that involves three phases:
modelling, experimentation, and optimization (Shi and
AbouRizk, 1997). It is praised for its ability to study
a real world system in any desired detail and its ability
to address the random and dynamic features in the
operation of the system. However, the complexity
involved in modelling a real world problem and in
experimenting with the model greatly increases the

Construction Management and Economics

ISSN 01446193 print/ISSN 1466-433X online 2002 Taylor & Francis Ltd
http://www.tandf.co.uk/journals
DOI: 10.1080/01446190210151032

Shi

484
chances of getting invalid results that do not characterize the system. Generally speaking, there are three
types of error in simulation (Pidd, 1998).
l

Downloaded by [Univ Autonoma De Yucatan] at 18:03 14 March 2014

Type zero errors. This type of error occurs when

the modeller asks the wrong questions, so that
the model does the totally wrong thing, or the
model does not operate in the manner in which
it is intended. Mostly this type of error occurs
if the modeller does not have a good understanding of the real world system and/or the
issues concerned with the system.
Types I and II errors. Type I errors correspond
to classical statistical hypothesis testing when a
valid model is wrongly rejected because there is
a certain probability that an error may occur.
Type II errors occur when a false model is
accepted as valid because of the accuracy of
statistics.

Types I and II errors cannot be avoided but can be

estimated by a con dence level. Type zero errors are
severe and must be avoided. To ensure a simulation
free of type zero errors, it is necessary to verify that
the simulation model does characterize the real world
system and that the simulation results obtained are a
valid representation of the systems performance. In
Zeiglers (1984) terms, veri cation is a process to
assure the simulation model is properly realized, and
validation is a process to assess the degree to which
the simulation model inputoutput relations map onto
those of the system. Two distinct approaches are used
for simulation veri cation and validation: black-box
and white-box approaches (Pidd, 1998). As the name
suggests, the black-box approach treats a real world
process and its corresponding simulation model as two
separate black boxes. The same set of parameters is
selected for measuring the performance of the process
and the model, and the two sets of data obtained are
tested for difference. The basic hypothesis is that the
data obtained from the model should be indistinguishable from those of the process if a simulation is
valid. Details of the methodology for using hypothesis
tests in comparing a models and a systems output
data for model validation are given in Balci and Sargent
(1981).
A simulation usually can provide a comprehensive
set of required results. However, validating a simulation using a black-box method still faces two challenges: (a) selecting measure parameters, and (b)
collecting data for the real world process. A process
can always be measured by different parameters. For
instance, usually a construction process is measured by
its production rate, operations of activities, and utilization of key resources. Selecting a different set of
parameters may result in a different conclusion on

whether or not to accept a simulation. After parameters

are selected, the remaining challenge is to obtain the
data observed for the process. Unfortunately, these
data are not always available. A simulation would not
be needed if these actual data were indeed available.
Although a black-box method provides a straightforward approach for validating a simulation study, its
applicability may be limited for two reasons: (a) a simulation may not be validated if selected parameters are
not comprehensive, and/or (b) actual data may not be
available for a real world process. When actual data
are not available for a system, AbouRizk et al. (1991)
suggested using other methods for estimating the
values of selected parameters of the system. The results
obtained are then compared with those obtained from
simulation. The method productivity delay model
(MPDM; Adrian and Boyer, 1976) and queuing theory
were used for validating the simulation results of a roof
truss installation process. Productivity was selected as
the parameter. The authors argued that the productivity values obtained from the three methods should
be consistent under the same given initial conditions.
The black-box approach af rms whether the simulation results are in a reasonable range, so that a user
can establish con dence in the obtained results. However, the approach does not eliminate the possibility
of errors in a simulation, either in a simulation model
or its experiments, or in both. In other words, close
results may be coincidently obtained from a simulation
even if it is not correctly conducted.
To ensure the validity of a simulation, the simulated
behaviour of the model must be in line or be comparable with the actual performance of the process in the
real world. In other words, the model must operate in
the manner intended. This study presents three easyto-use white-box methods for examining a dynamic
simulation process so as to verify and validate the simulation. These methods are implemented in the activitybased construction simulation system, or ABC (Shi
1999), in the form of reports. Therefore, validating
a simulation is achieved by selecting and validating
corresponding reports.

A brief description of ABC modelling and

simulation algorithm
ABC is an activity network-based modelling and simulation method using one single element, i.e. activity.
Various functions needed for modelling a construction
process are achieved by allowing activities to carry
attributes like activity duration, resources, logical
sequence, operational conditions, etc. The major difference between ABC and existing simulation systems
lies in the fact that existing simulation systems require

Downloaded by [Univ Autonoma De Yucatan] at 18:03 14 March 2014

Veri cation and validation of construction simulation

485

multiple elements from ve to three dozen, each of

which models one speci c function like resource
waiting and activity. The major bene ts of ABC are
ease of use and resultant time savings in constructing
simulation models. A study (Shi, 2000) has concluded
that ABC would generate equivalent results to other
construction simulation systems for the same process.
The ABC simulation assigns a unique identi cation
(ID) number to each resource entity when it is rst
released from its initial source (either an activity or the
resource pool); the entity is then identi ed by the ID
# for its entire lifecycle during simulation. Also its
movement in the simulation model and change of state
are traced chronologically along simulation time. The
ABC system traces an entire simulation progress in the
format:

is available for emptying the bucket and spreading the

concrete. The initial resource capacities are listed in
Table 1. To be consistent with the example for readers
to compare the results, deterministic values are directly
used from the book, although random distributions
can be used in ABC. Simulating the process with
MicroCyclone, the production rate is ten spreading
cycles per hour for the crew (Halpin and Riggs, 1992).
The corresponding ABC model is depicted in Figure
1, in which the required truck unit at activity 1 is 5
representing that a truck requires ve loading passes to
ll its ve compartments. Running the model in the
ABC system, 50 production cycles are collected at
activity 10 (spread concrete) in the total simulation
time of 304 minutes, which is equivalent to the production rate obtained from MicroCyclone.

Tnow AL AN FT D R/ID(IT)
Here Tnow = current simulation time; AL = activity
label; AN = activity name; FT = nish time; D = activity duration; and R/ID(IT) = resource name/identi cation number (idle time).
Each line is recorded right after an activity is selected
to start its operation. It details when the activity starts
and nishes its construction, as well as the information on the resources involved. An example line would
be:
25 1 load_truck 30 5 tower/1(0), truck/3(25)
This line shows: simulation time is advanced to time
25; activity 1 load truck starts operation; it takes ve
minutes to operate, and completes at time 30; two
resources (tower and truck) are engaged in the operation of the activity. The tower entity is labelled #1
and has zero waiting time before the activity starts.
The truck entity is labelled #3 and has waited 25
minutes before the activity starts its operation.

A concreting process
To illustrate the methods, a concreting process from
Halpin and Riggs (1992) is used. The reasons for
selecting this process are: (a) it is a construction
process common in building and road construction,
and (b) it represents a reasonable level of complexity
needed for modelling and simulation. In this process,
open-bay trucks with ve compartments are batched
with ve dry batches of concrete (one compartment at
one time) at the batch tower. The loaded truck hauls
the concrete to a pavement job site, and dumps the
batches separately into the skipper of a mixer near the
paving site. The mixer then mixes the concrete and
dumps the wet concrete into a bucket. The bucket is
then lifted to the pavement location by a crane. A crew

White-box methods for verifying and

validating a simulation
Many white-box methods are available for verifying
and validating a simulation (Sargent, 2000), with the
general idea of examining the internal structure of a
simulation model or looking into how a simulation
experiment is conducted. A simulation process involves
extensive logical and mathematical computations, and
deals with issues like simulation entities, activities, and
events over time. Understanding how a simulation is
conducted requires comprehensive knowledge and
Table 1

Resource capacity

Resource name

Capacity

(1)

(2)

Batch tower
Trucks
Mixer
Buckets
Crane
Crew

Figure 1

1
4
1
2
1
1

The ABC model of the concreting process

Shi

Downloaded by [Univ Autonoma De Yucatan] at 18:03 14 March 2014

486
experience. As a result, white-box methods can be
mastered only by advanced users.
From a construction engineers viewpoint, a real
world construction process is characterized by operations of activities. An activitys operation is observable
and measurable in terms of its start and nish times,
the sequence of its execution, and the resources
engaged. A valid model must be able to simulate the
operations of all activities in the manner in which they
would operate in the real world. A simulation is valid
if the simulated operations are identical or equivalent
to the expected actual operations; otherwise it is
invalid.
Three activity-based white-box methods are discussed here: (a) the chronological order of activity executions, (b) operations and durations of activities, and
(c) activity cycles of resource entities.

shown that it is a time-consuming and tedious task to

examine a long tracing report. Animation technology
provides an effective means for a user to visualize a
simulation progress on a screen so that it may be veri ed visually (Zhang and Shi, 2002). However, verifying the following issues provides a user with a quick
diagnosis on a simulation model and its operation.
Starting activities
A simulation must start with the correct activities as
expected. For example, an earthmoving process should
start with activities for truck(s) being loaded with soil.
A quick examination of the beginning lines in a tracing
report would assist a user in answering the key questions:
l
l

The chronological order of activity executions

All activities must be executed in the right logical order

in a simulation. When Activity X is examined for its
execution sequence, it must pass two tests.
Logical test: this test veri es whether all predecessors of activity X are completed before activity X starts.
Time test: this test validates the start time of
activity X. An activitys start time is constrained
by at least two factors: preceding logical, and
resources de ned by the following equations

S x = max {Fj|j = predecessor}

(1)

(2)

= max {Rk|k = resource}

Sx = max {S x,S x }

An activitys start is constrained by a series of conditions such as resources, simulation entities, and preceding logic. It cannot be scheduled for start until all
required conditions are met. If no activity can be
scheduled for start, a simulation would result in an
empty tracing report. If this is happening in a simulation, a user should check the start conditions of the
activities, especially those that are expected to start at
the beginning of the simulation against the initial conditions. When a simulation cannot start, this is mostly
due to the following two reasons.
l

(3)

Here Fj = the nish time of predecessor j; Rk = the

available time of resource k; S x = the earliest feasible
start time allowed by predecessors; S x = the earliest
feasible start time allowed by resources; Sx = the earliest feasible start time of activity X.
The operation of an activity is activated by simulation entities. Con rming the sequence of an activitys
execution would prove that the simulation entities ow
in the right logical order during simulation. Otherwise,
if some activities were executed in the wrong order,
the simulation model would contain errors causing
simulation entities to move in the wrong directions.
Listing all the activities in the chronological order of
their executions greatly assists a user in carrying out
the two tests for a simulation. A user then must
conduct the tests for all the activities on the list. It is
common to see a list containing hundreds or thousands of lines. In a repetitive process, many activities
repeat their operations over time. Experience has

which activities have started when a simulation

begins?
are those starting activities expected?
are any activities expected to start but have not
started?

Unneeded conditions: if an activity is assigned

with unneeded conditions at the modelling
stage, it would not be able to start because such
conditions cannot be met.
Problems related to initialization: usually, starting activities are initialized with their needed
resources and other conditions at the modelling
stage.

The rst-round operation of activities

It is meaningful to examine the rst operating cycle of
a model. Such an examination con rms whether activities are started and are advanced in the correct
sequence, and whether resource entities are initialized
at the right locations. Any inconsistencies indicate
errors in the model. If a cyclic process operates in the
right manner in the rst cycle, the remaining repetitive cycles probably would operate in the same manner,
although this may not always be true. For instance,
the activities in an earthmoving process must follow
the order in its rst cycle, i.e. Load_truck, Haul_to,
Dump_truck, and Truck_return.

Veri cation and validation of construction simulation

Last executed activities
If a simulation was terminated as expected, the report
contains all the activities executed in the simulated
time period. However, it may terminate before reaching its target if a model has errors. Examining last executed activities would assist a user in diagnosing why
the simulation could not proceed from that time point,
so that errors in the model could be detected.

Downloaded by [Univ Autonoma De Yucatan] at 18:03 14 March 2014

ABC tracing reports

Tracing a simulation process is not new. A tracing
report is available in many simulation systems such as
SLAMSYSTEM (Pritsker et al., 1997). However,
usually it contains all the events and elements executed
during simulation, so the report is very long and it is
not straightforward for engineers to understand. An
ABC tracing report contains only activities listed in the
chronological order of their executions. This feature
makes the ABC tracing report unique and easier to
understand.
An example tracing report for the concreting process
is shown in Table 2, which depicts that the simulation
starts off with a truck (ID# 2) being loaded with ve
batches of dry concrete. While the truck is fully loaded
at time 25, the next truck (ID #3) starts being loaded
and the loaded truck (ID# 2) starts travelling to the
mixer (activity 2). When simulation time is advanced
to 30, truck ID# 3 is still under loading with dry concrete. When time is advanced to 35, truck ID# 2
arrives at the mixer and starts dumping the dry concrete to the skipper, and it takes one minute for
dumping. The mixer starts mixing at time 36. The
mixed concrete is ready after three minutes and is
dumped to a bucket at time 39 . . . Following the simulation time in the report, a user can examine the
sequence of activities executed. Obviously, the progress
must be equivalent to the actual construction progress.
Operating counts and durations of activities
An activitys operation can be measured by two statistics in a simulation: operating count (the number of
operations that an activity has executed in the simulated time period), and mean duration (the average
time for executing the activity). Often the production
of a construction process is calculated from counting
the number of operations of a production activity. For
instance, the production of an earthmoving process is
measured by the number of truck loads of soil dumped
(i.e. number of operations of the activity dump). A
production activity must collect the right number of
operations in order to validate the processs production; otherwise, the model has errors probably related
to the production activity.

487
Commonly an activitys duration is described by a
stochastic distribution in simulation. A value is sampled from the given distribution function as the
activitys duration each time before its execution.
Usually, the values are different in different operations. Sometimes the sampled values may be biased,
especially when the activity has registered a small number of operations. Moreover, incorrect entries of a distribution function signi cantly affect a simulation. The
mean activity duration would give a user the idea of the
signi cance of the speci cation. For instance, if the
loading time in the concreting example is speci ed with
a uniform distribution in the range from two to seven
minutes instead of a constant value of ve minutes, the
simulation report shows a mean duration of 4.36 minutes collected from 50 loading operations. Is this
expected? If not, a correction is needed for the
activitys duration.
The ABC system generates a summary report as
shown in Table 3 for the concreting process. The operating counts of all activities are listed in column 3,
which shows that activities 2 and 4 have operated 10
times and all other activities have operated 50 times.
Because each truck requires ve loads to ll its box,
and each truck load can feed the mixer ve times, the
operating counts are consistent. The mean activity
durations obtained equal the speci ed deterministic
durations, and are listed in column 4 of Table 3.
Any inconsistency in operating counts among activities indicates errors in a simulation model. For
example, if the ABC model of the concreting process
is not correctly connected, as shown in Figure 2, in
which one arrow links activity 9 to activity 6 and
another arrow links activity 9 to activity 7 (dashed
lines). Subjectively, the user may expect bucket entities to be returned to activity 6, and crane entities to
be routed to activity 7. With the modi ed relationships, we run the model again. The operating counts
of activities are listed in column 5 of Table 3, which
shows that activities 7 to 10 have operated 84 times,
although they should have operated only 50 times. A
close look reveals the problem. Because of the two
added connections, the resource entity bucket shortcircuits to activity 7 instead of activity 6 as expected.
If the two connections are deleted, both crane and
bucket entities are released to the resource pool, from
which bucket is assigned to activity 6 and crane is
released to activity 7, respectively.
Activity cycles of resource entities
A resource entity interacts with one or more activities
during simulation. If all engaged operations of a
re-source entity are singled out and are arranged in
a chronological order, corresponding activities may
constitute one of the following patterns.

Shi

488

Downloaded by [Univ Autonoma De Yucatan] at 18:03 14 March 2014

Table 2

Simulation progress report

Tnow

Act label

Act name

Finish time

Duration

0
5
10
15
20
25
25
30
35
35
36
39
39.5
39.5
39.75
40
40.05
40.05
40.5
43.5
44
44
45
45
45.05
45.35
45.35
48
48.5
48.5
49.5
50
50

1
1
1
1
1
1
2
1
1
3
5
6
3
7
8
1
9
10
5
6
3
7
5
1
8
9
10
6
3
7
5
1
2

Load_truck
Load_truck
Load_truck
Load_truck
Load_truck
Load_truck
Travel_to_mixer
Load_truck
Load_truck
Dump_to_skip
Mix
Fill_bucket
Dump_to_skip
Swing_crane
Empty_bucket
Load_truck
Crane_return
Spread_concrete
Mix
Fill_bucket
Dump_to_skip
Swing_crane
Mix
Load_truck
Empty_bucket
Crane_return
Spread_concrete
Fill_bucket
Dump_to_skip
Swing_crane
Mix
Load_truck
Travel_to_mixer

5
10
15
20
25
30
35
35
40
36
39
39.5
40.5
39.75
40.05
45
40.25
45.05
43.5
44
45
44.25
48
50
45.35
45.55
50.35
48.5
49.5
48.75
52.5
55
60

5
5
5
5
5
5
10
5
5
1
3
0.5
1
0.25
0.3
5
0.2
5
3
0.5
1
0.25
3
5
0.3
0.2
5
0.5
1
0.25
3
5
10

One single activity: if the resource is devoted to

one single activity, it would only engage in this
activity while its state changes between busy and
idle during simulation.
A simple repetitive cycle: the activities may constitute a simple repetitive cycle. In this case, the
entity starts from the rst activity, moves
forward to other activities until the last one in
the cycle by following the logical order; it starts
again from the rst activity and repeats the same
cycle over time.
Non-repetitive cycles: these activities may form
multiple cycles, and a resource entity may take
one cycle in one round and may take a different one in the next round.
An arbitrary collection of activities: although a
resource entity may engage in certain activities
during simulation, these activities do not form
any cycle or pattern.

Resource
Tower/1(0),Truck/2(0)
Tower/1(0),Truck/2(0)
Tower/1(0),Truck/2(0)
Tower/1(0),Truck/2(0)
Tower/1(0),Truck/2(0)
Tower/1(0),Truck/3(25)
Truck/2(0)
Tower/1(0),Truck/3(0)
Tower/1(0),Truck/3(0)
Mixer/4(35),Truck/2(0)
Mixer/4(0)
Mixer/4(0),bucket/5(39)
Mixer/4(0),Truck/2(3.5)
Crane/6(39.5),bucket/5(0)
Bucket/5(0),crane/6(0),labor/7(39.75)
Tower/1(0),Truck/3(0)
Crane/6(0),bucket/5(0)
Labor/7(0)
Mixer/4(0)
Mixer/4(0),bucket/8(43.5)
Mixer/4(0),truck/2(3.5)
Crane/6(3.75),bucket/8(0)
Mixer/4(0)
Tower/1(0),Truck/3(0)
Bucket/8(0.79),crane/6(0.79),labor/7(0)
Crane/6(0),bucket/8(0)
Labor/7(0)
Mixer/4(0),bucket/5(7.75)
Mixer/4(0),truck/2(3.5)
Crane/6(2.95),bucket/5(0)
Mixer/4(0)
Tower/1(0),Truck/9(50)
Truck/3(0)

In general, a resource entity r ows from activity X

to activity Y. The entity must pass two tests in order
to validate its operation:
l

logical test-activity Y must be a correct successor for the resource entity to enter. Y = X
if the entity stays at the activity;
time test-the entity must pass through various
events at correct times.

A resource entity goes through three events at each

activity, as depicted in Figure 3: entering an activity,
starting the operation of the activity, and leaving the
activity (entering the next activity). Activity Xs start
time (Sx ) must be equal to the sum of the entitys entering time (Exr ) and its waiting time (Wxr ) at the activity
as de ned in Eq. 4. Activity Xs nish time (Fx ) equals
the activitys start time plus its duration (Dx ), and the
time that entity r enters into activity Y must be equal
to activity Xs nish time as shown in Eq. 6. A

Veri cation and validation of construction simulation

Table 3
Activity #

(1)

Downloaded by [Univ Autonoma De Yucatan] at 18:03 14 March 2014

1
2
3
4
5
6
7
8
9
10

489

Operating counts of activities

Activity name

(2)

The original model

Operating counts of
the modi ed model

Counts
(3)

Mean dur.
(4)

(5)

50
10
50
10
50
50
50
50
50
50

5
10
1
8
3
0.5
0.25
0.3
0.2
5

50
10
50
10
50
50
84
84
84
84

Load truck
Travel to mixer
Dump to skip
Truck return
Mix
Fill bucket
Swing crane
Empty bucket
Crane/bucket return
Spread concrete

in Table 4. The entity starts off at activity 6 ( ll

bucketSeq. #1), succeeded by activity 7 (swing
craneSeq. #2), then activity 8 (empty bucketSeq.
#3), and next activity 9 (crane returnSeq. #4). After
that, the entity is routed to activity 7 (swing craneSeq.
#5) instead of activity 6 ( ll bucket) as it should be.
The cyclic report indicates clearly where the shortcircuiting occurs, so that we should remove the two
links added in the modi ed model in Figure 2.
Figure 2 The modi ed ABC model of the concreting
process

Figure 3

An entity owing between two activities

simulation must pass the two tests for all resource entities in order to be valid; otherwise it is invalid.
Sx = E xr + W xr

(4)

Fx = Sx + Dx

(5)

E Yr = Fx

(6)

To obtain the cyclic report of a resource entity, its

entire lifecycle must be traced in simulation. ABC
traces and generates a cyclic report for each resource
entity. To obtain a report, a resource and a unit must
be selected if the resource has multiple units. The
cyclic report lists all activities in a chronological order
that the selected entity has engaged during simulation.
The cyclic report of bucket #1 for the modi ed
model (Figure 2) of the concreting process is shown

Table 4

Cyclic report of bucket 1

Seq No.

Act. name

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25

Fill_bucket
Swing_crane
Empty_bucket
Crane_return
Swing_crane
Empty_bucket
Crane_return
Fill_bucket
Swing_crane
Empty_bucket
Crane_return
Fill_bicket
Swing_crane
Empty_bucket
Crane_return
Fill_bucket
Swing_crane
Empty_bucket
Crane_return
Fill_bucket
Swing_crane
Empty_bucket
Crane_return
Fill_bucket
Swing_crane

Idle time

Start

Finish

39
0
0
0
0
4.54
0
2.45
2.34
4.54
0
0.85
3.94
4.55
0
1.75
3.05
4.55
0
0.14
4.65
4.55
0
1.04
3.75

39
39.5
39.75
40.05
40.25
45.05
45.35
48
50.85
55.65
55.95
57
61.45
66.25
66.55
68.5
72.05
76.85
87.75
77.5
82.65
87.45
87.75
89
93.25

39.5
39.75
40.05
40.25
40.5
45.35
45.55
48.5
51.1
55.95
56.15
57.5
61.7
66.55
66.75
69
72.3
77.15
87.95
78
82.9
87.75
87.95
89.5
93.5

490

Downloaded by [Univ Autonoma De Yucatan] at 18:03 14 March 2014

Conclusion
Simulation provides a quantitative approach for studying a real world system based on a mathematical/logical
model. Only valid, credible and correct simulation
results are acceptable and useful for improving a
systems performance. The black-box approach is
helpful for assisting users in determining whether the
simulation results obtained are in a reasonable range
by comparing them with the actual observations of the
actual system or the results derived from any analytical methods. On the other hand, the white-box
approach provides a means for users to look inside how
a simulation experiment is conducted so as to validate
the results obtained.
Activities characterize a construction process. Their
operations are observable and measurable in the real
world. This research studied how a construction simulation can be veri ed and validated by examining
three categories of simulation results about activities:
(a) the chronological order in which activities are conducted, (b) the operating counts and durations of activities, and (c) the activity cycles of resource entities.
Integrating with the ABC simulation system, the three
validating approaches are implemented as three simulation reports. Jointly con rming these reports, users
can effectively debug their simulation models and validate the simulation results.
The methods presented have been used mainly in
classrooms by graduate students for solving their
assignments and developing their term projects originated from real world construction processes. Together
with the ABC system, users can obtain valid simulation results in much less time. Limited comparative
analyses showed that a simulation project consumes
one-third to one-half of the time usually needed by
using other simulation systems. The ABC system is in
its second year of testing. Interested readers may
contact the author for an examination copy.

References
AbouRizk, S. M., Halpin, D. W. and Hill, S. L. (1991)
Measuring productivity and validating microcomputer simulation. Microcomputers in Civil Engineering, 6(2), 20215.
Adrian, J. and Boyer, L. T. (1976) Modeling method-productivity. Journal of the Construction Division ASCE, 103(2),
15468.
Balci, O. and Sargent, R. G. (1981) A methodology for costrisk analysis in the statistical validation of simulation
models. Communications of the ACM, 24(4), 1907.
Chang, D. (1987) Resque. Ph.D. thesis, University of
Michigan.
Hajjar, D. and AbouRizk, S. M. (1999) Simphony: an environment for building special purpose construction simula-

Shi
tion tools, in Proceedings of the Winter IEEE Simulation
Conference, IEEE, Piscataway, NJ, pp. 9981006.
Halpin, D. W. (1977) Cyclone: a method for modelling job
site processes. Journal of Construction Division ASCE,
103(3), 48999.
Halpin, D. W. and Martinez, L. H. (1999) Real world applications of construction process simulation, in Proceedings
of the Winter IEEE Simulation Conference, IEEE, Piscataway, NJ, pp. 95662.
Halpin, D. W. and Riggs, L. S. (1992) Planning and Analysis
of Construction Operations, Wiley, New York.
Huang, R., Grigoriadis, A. M. and Halpin, D. W. (1994)
Simulation of cable-stayed bridges using Disco, in
Proceedings of Winter IEEE Simulation Conference, IEEE,
Piscataway, NJ, pp. 11306.
Ioannou, P. G. (1989) UM-Cyclone Users Guide. Department
of Civil Engineering, The University of Michigan, Ann
Arbor, MI.
Liu, L. Y. and Ioannou, P. G. (1992) Graphical object-oriented discrete-event simulation system, in Proceedings of
1992 Winter IEEE Simulation Conference, IEEE, Piscataway,
NJ, pp. 128591.
Martinez, J. C. and Ioannou, P. G. (1999) General purpose
systems for effective construction simulation. Journal of
Construction Engineering and Management ASCE, 125(4),
26576.
Paul, R. J. (1991) Recent developments in simulation modelling. Journal of the Operational Research Society, 42(3),
21726.
Paulson Jr., B. C. (1978) Interactive graphics for simulating
construction operations. Journal of the Construction Division
ASCE, 104(1), 6976.
Pidd, M. (1998) Computer simulation for operational
research in 1984, in Developments in Operational Research,
Eglese, R. W. and Rand, G. K. (eds), Pergamon Press,
Oxford.
Pidd, M. (1992) Computer Simulation in Management Sciences,
3rd Edn, Wiley, Chichester.
Pritsker, A. A. B., OReilly, J. J. and LaVall, D. K. (1997)
Introduction to Simulation and SLAM-II, 2nd Edn, Wiley,
New York.
Robinson, S. W. (1996) Service quality management in
the process of delivering a simulation study. Paper presented at the 14th Triennial Conference of the
Inter-national Federation of OR Societies, 812 July,
Vancouver, BC.
Sargent, R. G. (2000) Veri cation, validation, and accreditation of simulation models, in Proceedings of the Winter
IEEE Simulation Conference, IEEE, Piscataway, NJ,
pp. 509.
Shi, J. and AbouRizk, S. M. (1997) Resource-based modeling for construction simulation. Journal of Construction
Engineering and Management ASCE, 123(1), 2633.
Shi, J. (1999) Activity-based construction (ABC) modeling
and simulation method. Journal of Construction Engineering
and Management ASCE, 125(5), 35460.
Shi, J. (2000) Object-oriented technology for enhancing
activity-based modeling functionality, in Proceeding of the
2000 Winter IEEE Simulation Conference, IEEE, Piscataway,
NJ.

Downloaded by [Univ Autonoma De Yucatan] at 18:03 14 March 2014

Veri cation and validation of construction simulation

491

Teicholz, P. (1963) A Simulation Approach to the Selection

of Construction Equipment, Technical Report No. 26,
The Construction Institute, Stanford University, Palo Alto,
CA.
Tommelein, I. D., and Odeh, A. M. (1994) Knowledgebased assembly of simulation networks using construction
designs, plans, and methods, in Proceedings of the 1994

Winter IEEE Simulation Conference, IEEE, Piscataway, NJ,

pp. 114558.
Zeigler, B. P. (1984) Multifaceted Modelling and Discrete Event
Simulation, Academic Press, London.
Zhang, H. and Shi, J. (2002) Iconic animation for activitybased construction simulation. Journal of Computing in Civil
Engineering ASCE, 16(3), 15764.