Professional Documents
Culture Documents
HUANG Jian-wen
College of Economic & Management
China Three Gorges University
Yichang, P.R.China
javenhuang@126.com
WANG Xing-xia
College of Civil & Hydroelectric Engineering
China Three Gorges University
Yichang, P.R.China
wangxingxia1980@126.com
AbstractIn a construction project where duration truly
equals money, the management of duration is critical, thus
predicting the likelihood of duration may play a key role
towards project success. On the basis of the researches of
other scholars, this paper analyzed the traditional probability
analysis method for duration risk in program evaluation and
review technique(PERT). On basis of that it simulates the
projects duration and analyzes the risk of construction
schedule by Monte Carlo simulation(MCS) method. Monte
Carlo method was used to simulate the duration for each
activity and the overall project to acurately determine the
completion probability of the project under considering of the
changeability and randomness of duration for each activity.
The result shows that the MCS method is convenient, effective
and efficient. It provides scientific basis for project decision-
making and supports the decider powerfully.
Keywords-construction schedule; probability distribution;
duration risk; Monte Carlo simulation(MCS)
I. INTRODUCTION
In this worldwide competition, regarding limited
resources, project management is paid increasing attention.
Planning and decision making are the basic duties of a
project manager and without them it is impossible to reach
project goals[1]. Construction projects are one-off endeavors
with many unique features such as long period, complicated
processes, abominable environment, financial intensity and
dynamic organization structures and such organizational and
technological complexity generates enormous risks.
Schedules are essential to the successful execution of
projects. It is widely accepted that construction project
schedule plays a key role in project management due to its
influence on project success[2~6].
The PERT(Program Evaluation and Review Technique)
is widely used as a tool for managing large-scale projects. It
was introduced in the 1950s to estimate the time required for
a given task[7]. During the 1960s and 1970s, various authors
proposed a revision of the hypotheses put forward by the
creators of the PERT method and several authors have
modified the original three point PERT estimators to
improve the accuracy of the estimates[3]. Eugene David
Hahn[2] argues that the tail-area decay of an activity time
distribution is a key factor which has been insufficiently
considered previously and provides a distribution which
permits varying amounts of dispersion and greater
likelihoods of more extreme tail-area events that is
straightforward to implement with expert judgments.
Monte Carlo analysis is a statistical technique that could
become increasingly important as a means for risk assessors
to evaluate the uncertainty. Although Monte Carlo
simulation has been used since the 1940s, more powerful
desktop computers have made it accessible and attractive for
many new applications. In [1], Monte Carlo simulation
analysis is implemented as an integrated tool to reach the
project goals, analyzing and investigating a variety of
uncertainty permutations simultaneously. Tsuen-Ho Hsu and
Frank F.C. Pan[9] extends the applications of both AHP and
the Monte Carlo simulation in service industry management
based on examining the quality structure of dental services.
In [10],Yun-Fu Wu introduces the correlated sampling
techniques to incorporate with conventional Monte Carlo
simulation to save engineers work as well as computing
time. PAN Chun-guang[11] builds a programming model to
find the critical path in fuzzy planning network and the
model can be transformed into a multiple objective linear
programming. Marta Schuhmacher[12] uses Monte Carlo
simulations techniques in risk assessment and the result
shows that the estimate of the coefficient of relative risk
aversion tends to have a negative bias. Using Monte Carlo
simulation, Miguel A. Usbel[13] obtains an unbiased and
consistent point estimator, and also confidence intervals as
approximations of a special case of multiple integral
frequently used in risk theory. In a word, Monte Carlo
simulation is a viable tool to provide numerical estimations
of the stochastic features of the system response[14,15].
II. TRADITIONAL PROBABILITY ANALYSIS METHOD
Duration risk means the possibility and loss of
incompletion in the total stipulated duration limit. According
to the definition of duration risk, the mathematical
expression of duration risk can be defined as follows:
( , )
f
R f p c = (1)
in which R is duration risk, pf is probability of incompletion
in the total stipulated duration limit, and c is loss for
incompletion.
When the total stipulated duration limit is Ts, the
completion probability can be defined by formula (2).
( ) ( )
s
T
s
p T T f t dt
(2)
in which ( ) f t is the probability density function of projects
total duration[4,5,15].
2009 International Conference on Information Management, Innovation Management and Industrial Engineering
978-0-7695-3876-1/09 $25.00 2009 IEEE
DOI 10.1109/ICIII.2009.195
150
2009 International Conference on Information Management, Innovation Management and Industrial Engineering
978-0-7695-3876-1/09 $25.00 2009 IEEE
DOI 10.1109/ICIII.2009.195
150
2009 International Conference on Information Management, Innovation Management and Industrial Engineering
978-0-7695-3876-1/09 $26.00 2009 IEEE
DOI 10.1109/ICIII.2009.195
150
PERT assumes that the duration of each activity is a
random variable and it obeys Beta distribution[3,7]. The
probability density function of Beta distribution is as
follows:
1 2
1 2
1 1
1 2
1
1 2
( )( ) ( )
( )
( ) ( )( )
; , 0
k k
k k
k k t a b t
f t
k k b a
a t b a b
+
+
=
>
(3)
in which
1
0
( )
k x
k x e dx
+
=
.
For each activity i , we must estimate its optimistic
duration
i
a , pessimistic duration
i
b , and most likely
duration
i
m . Then we can calculate the mean
i
and the
variance
2
i
of duration for each activity by the following
formula:
4
6
i i i
i
a m b
+ +
=
,
2
2
( )
36
i i
i
b a
=
(4)
According to the central limit theorem, the sum of
independent identically distributed random variables is
approximately distributed normally. That means the project
duration T obeys normal distribution
2
( , ) N , so we can
calculate completion probability by following formula:
2
1 1
( ) exp[ ( ) ]
2 2
s
T
s
t
p T T dt
(5)
In the formula (5), the parameters and can be
calculated by following formula (4).
Because of
2
( , ) T N ,
(0,1)
T
N
, so we can
calculate completion probability by following formula:
2
1 1
( ) exp( )
2 2
s
T
s
p T T t dt
( )
s
T
=
(6)
Then we can determine completion probability p for the
total stipulated duration limit T
s
by calculating
s
T
X
=
and consulting standard normal distribution chart[16].
III. MONTE CARLO SIMULATION METHOD
A. Generation of pseudorandom numbers
In order to simulate the construction schedule risk, we
must take samples for the random variable, so it is
impossible to generate a sequence random numbers obeying
some distribution. The most commonly used technique to
generate a sequence pseudorandom numbers is called linear
congruential method[17]. Linear congruential method for
generating n random numbers is based on the fundamental
congruence relationship[8], which can be expressed as
following formula:
1
{ }(mod ), 0,1, 2,
n n
x rx c s n
+
= + = (7)
in which r is the multiplier, c is the increment, x
0
is the seed,
and s is an integer-valued modulus. The modulo notation
(mod s) represents that:
1 n n n
x rx c sI
+
= + (8)
with [( ) ]
n n
I rx c s = + denoting the largest positive integer
value in ( )
n
rx c s + .
Once the sequence of random number x is generated, the
random number from the unit interval [0,1]
n
can be
obtained as follows:
n
n
x
s
=
(9)
B. Process of Monte Carlo simulation
As far as risk analysis of construction schedule is
concerned, Monte Carlo simulation is performed in the
following steps[15,18~21]:
Step 1: Determine the distribution function for each
activity of the construction project and the simulation times.
Generally, the distribution of the activity can be normal
distribution, triangular distribution and Beta distribution. In
this paper we select Beta distribution as the distribution of
each activity.
Step 2: Estimate the optimistic duration a, pessimistic
duration b, and most likely duration m of each activity.
Step 3: Calculate the parameters k
1
and k
2
.
It is well known that the mean and the variance
2
of Beta distribution can be calculate as follows:
1
1 2
( )
k
a b a
k k
= +
+
,
2 1 2
2
1 2 1 2
( ) ( 1)
k k
k k k k
=
+ + +
(10)
But in PERT, the mean and the variance
2
of of
each activity can be estimated by formula (4). So we can get
equation (11) and (12) as follows.
1
1 2
4
( )
6
k a m b
a b a
k k
+ +
= +
+
(11)
2
1 2
2
1 2 1 2
( )
36 ( ) ( 1)
k k b a
k k k k
=
+ + +
(12)
On the other hand, about the mode m, because it is the
point of the maximum probability density it should meet
( )
0
df t
dt
=
. So we can calculate m as following formula:
151 151 151
1 2
1 2
( 1) ( 1)
( 2)
b k a k
m
k k
+
=
+
(13)
Then solving the simultaneous equations (11), (12) and
(13), we can get 3 solutions of the parameters k
1
and k
2
:
1 2
4 k k = = ,
1
3 2, k = +
2
3 2 k = and
1
3 2, k =
2
3 2 k = + .
Step 4: Generation of pseudorandom numbers
1
and
2
.
Step 5: Let
1 2
1 1
1 1
1
(1 )
( ( ) )
( )
k k
g a b a
b a
+ =
,
1 2
1 2
1 1
1
( ) ( )
( )
( )
k k
k k
m a b m
g m
b a
+
=
and calculate ( ) g m ,
1
( ( ) ) g a b a + and
1
( ( ) )
( )
g a b a
g m
+
.
Step 6: Compare with
1
( ( ) )
( )
g a b a
g m
+
and
2
. If
1
2
( ( ) )
( )
g a b a
g m
, let
1
( ) t a b a = + , else go to step
4.
Step 7: For each activity i , we can get its sampling
duration t
i
. Then we can calculate the duration T
of the
construction project by critical path method(CPM).
Step 8: Repeat step 4 to 7, generate the sequence {T
i
} of
the projects duration.
Step 9: Calculate the mean T and the standard variance
T
of the sequence {T
i
}.
Step 10: Calculate the completion probability by
2
1 1
( ) exp( )
2 2
s
T
T T
s
p T T t dt
.
IV. NUMERICAL EXAMPLE
Schedule planning of a certain project is shown as the
Figure.1 and the data of each activitys estimated duration
such as optimistic duration a, pessimistic duration b, and
most likely duration m are given in TABLE.I. If the total
stipulated duration limit T
s
is 104 days, now let us analyze
the completion probability of the project.
30,40,50
3
0
,3
6
,4
2
36,40,44 38,40,42 18,20,22
1
4
,2
0
,2
6
1
0
,2
0
,3
0
20,32,44
8
,1
0
,1
2
3
0
,4
0
,5
0
A
B
C
D
7
,1
0
,1
3
E
F
G
H
I
J
K
Figure 1. Net planning of project
TABLE I. ESTIMATED DURATION OF EACH ACTIVITY
Activity
(ij)
Estimated Duration
(days)
Activity
(ij)
Estimated Duration
(days)
a m b a m b
A(12) 30 36 42 G(46) 8 10 12
B(13) 36 40 44 H(47) 20 32 44
C(14) 30 40 50 I(58) 14 20 26
D(25) 30 40 50 J(68) 18 20 22
E(26) 7 10 13 K(78) 10 20 30
F(36) 38 40 42
A. Traditional probability analysis method
First, calculate mean duration
ij
and variance
2
ij
of
each activity of the project as TABLE.II showing.
TABLE II. CALCULATION RESULT OF EACH ACTIVITY
Activity
(ij)
Estimated Duration
(days)
ij
2
ij
ij
a m b
A(12) 30 36 42 36 4.0000 2.0000
B(13) 36 40 44 40 1.7778 1.3333
C(14) 30 40 50 40 11.1111 3.3333
D(25) 30 40 50 40 11.1111 3.3333
E(26) 7 10 13 10 1.0000 1.0000
F(36) 38 40 42 40 0.4444 0.6667
G(46) 8 10 12 10 0.4444 0.6667
H(47) 20 32 44 32 16.0000 4.0000
I(58) 14 20 26 20 4.0000 2.0000
J(68) 18 20 22 20 0.4444 0.6667
K(78) 10 20 30 20 11.1111 3.3333
Second, determine the critical path of the project, the
result is given in TABLE.III as follows.
TABLE III. DETERMINATION OF CRITICAL PATH
Feasible Path Total Mean Duration
1258 36+40+2096
1268 36+10+2066
1368 40+40+20100 Critical Path
1468 40+10+2070
1478 40+32+2092
Last, analyze completion probability of the project.
( , )
4
100
6
ij ij ij
i j A
a m b
+ +
= =
2
( , )
( )
1.6330
36
ij ij
i j A
b a
= =
2.4495
s
T
X
= =
So we can calculate the completion probability of the
project is ( ) (2.4495) 99.28%
s
p T T = . Although the
completion possibility is very high, the manager also should
pay more attention to it and strengthen risk control of
construction duration.
B. Monte Carlo simulation method
First, determine the distribution function for each activity
of the construction project as Beta distribution and the
simulation times=10000. Because there is ( ) 2 m a b = +
about the most likely duration of each activity, so we let
parameters k
1
=k
2
=4 in this example.
Second, generate pseudorandom numbers
1
and
2
by
uniform distribution in the interval [0,1]. Calculate
1
( ( ) )
( )
g a b a
g m
+
and Compare
2
with
1
( ( ) )
( )
g a b a
g m
+
.
In 10000 times simulation, we can find 4558 times
meeting
1
2
( ( ) )
( )
g a b a
g m