Beruflich Dokumente
Kultur Dokumente
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
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
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
485
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
Resource capacity
Resource name
Capacity
(1)
(2)
Batch tower
Trucks
Mixer
Buckets
Crane
Crew
Figure 1
1
4
1
2
1
1
Shi
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.
(1)
(2)
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)
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
Table 2
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
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)
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.
(1)
1
2
3
4
5
6
7
8
9
10
489
(2)
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
Figure 3
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)
Table 4
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
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.
491