IOP Conference Series: Materials Science and Engineering

PAPER • OPEN ACCESS

Comparison of PERT and M-PERT scheduling for a construction project

in Malang, Indonesia
To cite this article: Joshua Reinaldo Handoko and Onnyxiforus Gondokusumo 2019 IOP Conf. Ser.: Mater. Sci. Eng. 508 012034

View the article online for updates and enhancements.

This content was downloaded from IP address 181.215.10.71 on 03/05/2019 at 18:20

TICATE 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 508 (2019) 012034 doi:10.1088/1757-899X/508/1/012034

Comparison of PERT and M-PERT scheduling for a

construction project in Malang, Indonesia

Joshua Reinaldo Handoko1,* and Onnyxiforus Gondokusumo2

1
Post-graduate Student, Civil Engineering Department, Tarumanagara University, ID
2
Lecturer, Civil Engineering Department, Tarumanagara University, ID

*reinaldojosh2nd@gmail.com

Abstract. Construction projects planned using deterministic method suffers from delay
caused by uncertainties that are not considered in advance when determining activity
durations. Probabilistic method offers an alternative to consider such uncertainties using
PERT. Several studies on PERT however, proved that PERT results underestimated the
project duration mean and overestimated the project duration variance. In 2017, PERT was
developed into M-PERT. This study aims to compare project duration obtained from both
the PERT and M-PERT with project duration obtained from Monte Carlo simulation
software. Results of PERT and M-PERT on the same network model showed that M-PERT
produced smaller value of project duration variance compared to PERT at the same equal
value of project duration mean. M-PERT also produced project duration with RMSE value
towards the simulation results of 0.0001818. This value is smaller compared to PERT’s
RMSE value towards the simulation results of 0.0053961. This shows that M-PERT is able
to produce results that are closer to computer simulation results compared to PERT. This
study suggests that a probabilistic scheduling software which operates according to M-
PERT principles should be developed in order for M-PERT to be completely utilized in
actual construction projects.

1. Introduction
Project scheduling in construction projects commonly use deterministic method, where a certain value
is determined as the duration of the activity. This is done based on historical data from previous projects
or the scheduler’s judgement and experience. However, every construction project is unique from one
to another [1], meaning a successful plan from one project does not ensure success when applied in
another project. Another scheduling method that is less popular in project management is the
probabilistic method, where a project duration may be calculated when the activity duration is uncertain.
One of the probabilistic methods used in project management is the Program Evaluation and Review
Technique (PERT). It is used to calculate project duration from the optimistic, pessimistic and most
likely duration of the activities. However, PERT has its shortcoming in the form of result accuracy [2].
The cause of this problem is the network diagram modelling and no definitive solution for network
diagram without a dominant critical path [2], [3]. Ballesteros-Pérez [4] then reformulated PERT and
renamed it as M-PERT which allows more ways to model activity relationships to visualize their real-
life conditions in the network diagram. M-PERT calculates project duration by merging activities in the
network diagram repeatedly, providing a closer result to computer simulation compared to PERT [4],
[5]. This research is conducted by collecting time related data from an actual project and implementing
both PERT and M-PERT using that data to calculate the project duration. The results from both methods

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
TICATE 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 508 (2019) 012034 doi:10.1088/1757-899X/508/1/012034

are compared one another, and then compared with the results from Monte Carlo Simulation using a
computer software.

2. Method and materials

Research method used on this research are as follows:

2.1. Collection of Project Data

The required project data for this research was the activities in the project and their optimistic, most
likely and pessimistic duration. The type of project suitable for this research is a simple project such as
regular building with or less than 5 stories constructed in 2017 to 2018. List of activities were prepared
using the project time schedule, which contained information about all the activities that must be
completed and their planned completion time. This list of activities was then brought to the expert who
had experience and qualification in the project, to confirm how the activities were actually implemented
on site. The activities from the schedule may be divided into one or more detailed activities, or combined
into one major activity. The expert then suggested the optimistic, most likely and pessimistic duration
for each activity according to his or her experience on site.

2.2. Analysis Method

The three-point-estimate (optimistic, most likely and pessimistic duration) were used to calculate the
activity duration mean and variance. To calculate the project duration using PERT, the activity mean
previously calculated may be used as a single duration in the Arrow Diagram Network (ADN). After
completing the ADN, the critical path was identified. The project duration mean and variance for PERT
were calculated by directly adding up the mean and variance of the critical activities. As for M-PERT,
the activity variance must first be multiplied by correction factor K as explained in [6]. The network
diagram for M-PERT was modelled as Precedence Diagram Network (PDN) finish-to-start. Activity
relationships other than finish-to-start must be converted [7], [8]. Project duration mean and variance
for M-PERT were calculated through merging procedure where two or more activities in the network
are merged into a single activity with specific value of mean and variance. Merging procedure provided
in [4] was unique for every scenario in the network diagram [9], [10], [11]. Project duration mean and
variance from PERT and M-PERT were then compared with project duration mean and variance
obtained from Monte Carlo simulation to check their accuracy [12], [13], [14].

3. Results and discussion

Results obtained from this research were divided into the following sections:

3.1. Project Information

The selected project for this research was a 3 story supermarket building in Malang, East Java,
Indonesia. It was done from January 2017 to September 2017 for 38 weeks (266 days). The activities in
this project were divided into seven major categories : (1) preliminary works, (2) earthmoving works,
(3) concrete structure works, (4) MEP / utilities structure works, (5) steel structure works, (6)
architectural works and (7) external works.

3.2. Activity Duration Mean and Variance

The three-point-estimate obtained from the expert’s suggestions showed that in most activities, the
optimistic duration was the most likely duration. This is because the expert was confident that the
construction method implemented on site was sufficient to ensure the activity would be finished in the
alloted duration. Table 1 is a sample consisting of some activities in the project and their three-point-
estimate. The mean and variance for each activity were first calculated using PERT’s formula for mean
and variance. In accordance to the statement from Herrerías-Velasco et al. [6], the activity mean for M-
PERT is equal to PERT-calculated mean ; as for the activity variance, M-PERT required that the PERT-

2
TICATE 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 508 (2019) 012034 doi:10.1088/1757-899X/508/1/012034

calculated variance multiplied by a correction factor (K) if the scheduler decided to use three-point-
estimate. The K factor was also calculated using the optimistic, most likely and pessimistic duration.
Table 1. List of activities with their three point estimate, mean and variance
Act Optimistic Dur Realistic Dur Pessimistic Dur Estimated Mean PERT Variance Correction Factor M-PERT Variance
LIST OF ACTIVITIES Code O M P Te vTe K vTe MPERT
(weeks) (weeks) (weeks) (weeks) (weeks) (weeks)
PRELIMINARY WORKS

1.1 TEMPORARY FACILITIES A1 2 2 3 2.17 0.0278 0.7143 0.0198

EARTHMOVING WORKS B 6 6 8 6.33 0.1111 0.7143 0.0794
MAIN STRUCTURE

3.1 STRIP FOUNDATION + GROUND TIE BEAM C1 4 4 6 4.33 0.1111 0.7143 0.0794

3.2 GROUND FLOOR COLUMN C2 4 4 5 4.17 0.0278 0.7143 0.0198

3.3 GROUND FLOOR SLAB (ELEV ± 0.00) C3 2 3 5 3.17 0.2500 1.2222 0.3056

3.4 1ST FLOOR BEAM AND SLAB (ELEV + 4.95) C4 2 4 4 3.67 0.1111 0.7143 0.0794

3.5 1ST FLOOR COLUMN C5 2 3 4 3.00 0.1111 1.2857 0.1429

3.6 ROOF FLOOR BEAM AND SLAB (ELEV + 9.95) C6 2 3 5 3.17 0.2500 1.2222 0.3056
3.7 MEZANINE FLOOR COLUMN C7 2 2 4 2.33 0.1111 0.7143 0.0794
3.8 MEZANINE FLOOR BEAM AND SLAB (ELEV + 14.85) C8 2 2 4 2.33 0.1111 0.7143 0.0794

3.9 CONCRETE CONSOLE / LISTPLANK C9 1 1 2 1.17 0.0278 0.7143 0.0198

3.10 PARAPET C10 1 1 2 1.17 0.0278 0.7143 0.0198

3.11 STAIRS C11 2 4 5 3.83 0.2500 1.2222 0.3056

3.3. Project Duration Mean and Variance of PERT and M-PERT

The Arrow Diagram Network (ADN) for PERT is given in Figure 1 and Figure 2. Figure 1 is the first
half of the project consisting of preliminary works, earthmoving works and concrete structure works.
Figure 2 is the second half of the project consisting the MEP structure, steel structure, architectural
works, and external works. The blue, green and yellow node in Figure 1 will continue in the node with
the same colour in Figure 2. Notice that in these two figures, there are single activities that are divided
into two activities, for example activity C1. This is done to indicate that the next activity would start a
certain number of weeks after the current activity started. The activity mean (Te) as shown in Table 1 is
entered as the single duration of each activity. Using the CPM, the project duration was calculated and
the critical activities were determined. Critical activities were marked in Figure 1 and Figure 2 with the
yellow colour and red arrow. The project duration obtained from PERT was 38.34 weeks and the
variance was 2.7222 weeks, where both are the sum of critical activity mean and variance respectively.
0 A1 2,17 B (1) 4,17 B (2) 8,5
0 2,17 2M 4,17 4,33 M 11,3
2,17 M
C11

C1 (1) 5,17 C1 (2) 8,5 3,83 M E1 15,7

11,3 15,7
1M 5,17 3,33 M 4,33 M

C2 (1) 6,17 C2 (2) 9,34

1M 6,17 3,17 M 11,3

15,7
C4 (1) 7,17 C4 (2) 9,84 15,7

1M 7,17 2,67 M 11,3

C5 (1) 8,17 C5 (2) 10,2

1M 8,17 2M 11,3 C10

1,17 M
C6 (1) 9,17 C6 (2) 11,3

1M 9,17 2,17 M 11,3

C7 (1) 10,2 C7 (2) 11,5

1M 11,5 1,33 M 14,5

C8 (1) 11,2 C8 (2) 12,5 C9

12,5 1,33 M 14,5 1,17 M
1M

C3 (1) 12,2 C3 (2) 14,3

1M 13,5 2,17 M 15,7

Figure 1. PERT Arrow Diagram Network of case study project (part 1)

3
TICATE 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 508 (2019) 012034 doi:10.1088/1757-899X/508/1/012034

F1 20,7 F2
33,8
2,33 M 1,17 M

F3 (1) 16,7 F3 (2) 18,3 21.8 F6 25,2

15,7 E7 16,8 E5 18,3 30,2 31,5 35 38,3
1M 1,33 M 3,33 M
15,7 1,17 M 30,3 1,17 M 31,5
F7 23,8 F8
F4 18,8 F5 37,2
2M 1,17 M
E2 17 E4 32,7
2.17 M 2,33 M
1,33 M 30,2 1,33 M
15,7 G1 (1) 21,8 G1 (2) 23,8 G7 25,7
15,7 6,17 M 21,8 2M 24 1,83 M 28

G3

2M
G2 (1) 26 G2 (2) 28 G6 G11 36,3

4,17 M 26 28 6,5 M 1,83 M 38,3

2M
G4
3,33 M 34,5 G10 38,3
34,5 3,83 M 38,3

G5 27,8 G9
G8 21,8 G12 36,7
28 6,17 M
22,7 1,83 M 38,3
6,17 M 2,17 M

D12

12,2 C3 (2) 14,3 2,17 M 16,5 D13

13,5 15,7 24,8 H2 18 E3 19,3 E6
2,17 M 1,17 M
2,33 M 30,2 1,33 M 35,2 1,17 M
D1 15,2 D2 16,3 D8 17,5 D4
21,5 22,7 23,8 H1 20,3 H4 21,5 H3 24.2 H9
3M 1,17 M 1,17 M 2,17 M
2.33 M 34 1,17 M 35,2 1,17 M 36.3 2M

D9 17,5 D11 18,7

24,8 26 H5 21,5
1,17 M 1,17 M
1,17 M 35,2

D7 17,4 D5
24,8 H8 20,8 H6 23 H7
2,17 M 1,17 M
2,83 M 33 2,17 M 35,2 1,17 M

D6 13,3 D10 15,5

1,17 M 23,8 2,17 M 26

D3 15,7

2,33 M 27,8

Figure 2. PERT Arrow Diagram Network of case study project (part 2)

The Precedence Diagram Network (PDN) for M-PERT is given in Figure 3 and Figure 4. As the finish-
to-start model has been made when modelling PERT’s ADN, the layout for the M-PERT’s PDN model
is no different. Activities are written in the nodes instead of on the arrows. However, it is slightly
different than the usual PDN as in M-PERT, the variance of activity duration must also be considered
when calculating through merging procedure. Both mean and variance must be written in the activity
boxes as seen in Figure 3 and Figure 4. The node with the red coloured topside in Figure 3 and Figure
4 has the same function as the coloured node in Figure 1 and 2; to connect the two figures. Other things
that can be noticed from both figures are that division of single activities just like those in PERT’s ADN
must also be carried out and there is no critical path because the merging procedure do not require it.

START A B(i) B(ii) E1

2.17 0.02 2.00 0.00 4.33 0.08 4.33 1.22

C1(i) C1(ii) C1(iii) C11

1.00 0.00 3.33 0.08 0.00 0.00 3.83 0.31

C2(i) C2(ii) C2(iii)

1.00 0.00 3.17 0.02 0.00 0.00

C4(i) C4(ii) C4(iii)

1.00 0.00 2.67 0.08 0.00 0.00

C5(i) C5(ii) C5(iii) X3

1.00 0.00 2.00 0.14 0.00 0.00 0.00 0.00

C6(i) C6(ii) C6(iii) C10

1.00 0.00 2.17 0.31 0.00 0.00 1.17 0.02

C7(i) C7(ii) C7(iii)

1.00 0.00 1.33 0.08 0.00 0.00

C8(i) C8(ii) C8(iii) C9

1.00 0.00 1.33 0.08 0.00 0.00 1.17 0.02

C3(i) C3(ii) C3(iii) X1

1.00 0.00 1.33 0.31 0.00 0.00 0.00 0.00

Figure 3. M-PERT Precedence Diagram Network of case study project (part 1)

4
TICATE 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 508 (2019) 012034 doi:10.1088/1757-899X/508/1/012034

E1 E7 E5
4.33 1.22 1.17 0.02 1.17 0.02

E2 E4
1.33 0.08 1.33 0.08

F3 (i) F3 (ii) F1 F2 F6
1.00 0.00 1.33 0.08 2.33 0.08 1.17 0.02 3.33 0.08

F4 F5 F7 F8
2.17 0.02 2.33 0.08 2.00 0.14 1.17 0.02

X3 G1(i) G1(ii ) G7
0.00 0.00 6.17 0.31 2.00 0.00 1.83 0.02

FS+2
G3
2.00 0.14
G11
1.83 0.02
G2(i) G2(ii) G6
6.17 0.31 0.00 0.00 6.50 0.75
X4 G10 END
FS+2 0.00 0.00 3.83 0.31
G8 G4 G5 G9
6.17 0.31 3.33 0.08 FS -2 1.83 0.02 6.17 0.31
G12
FS+2 2.17 0.02
C3(i) C3(ii) C3(iii) X1 D12 D13 X2 1.00
0.00 1.33 0.31 0.00 0.00 0.00 0.00 2.17 0.02 1.17 0.02 0.00 0.00

D1 D2 D8 D4
3.00 0.14 1.17 0.02 1.17 0.02 2.17 0.02

D9 D11 E3 E6 H9
1.17 0.02 1.17 0.02 1.33 0.08 1.17 0.02 1.17 0.14

D7 D5 H1 H4 H3
2.17 0.02 1.17 0.02 2.33 0.08 1.17 0.02 1.17 0.02

D6 D10 H5
1.17 0.02 2.17 0.02 1.17 0.02

D3 H2 H8 H6 H7
2.33 0.08 2.33 0.08 2.83 0.02 2.17 0.02 1.17 0.02

Figure 4. M-PERT Precedence Diagram Network of case study project (part 2)

Merging procedure must be applied in order to calculate M-PERT project duration mean and variance.
In this case however, the project duration mean and variance could not be calculated manually. This was
due to the fact that at one point where the paths must be untangled into parallel paths (see illustration in
Figure 5), too many parallel paths were formed. As mentioned by Ballesteros-Pérez [4], correlation
factor must be calculated when there are tangled paths in the network that must be untangled. This means
even if somehow the scheduler was able to identify all the possible paths, he/she must still calculate the
correlation factor for all of that possible paths with a formula for manual calculation that was limited.
Ballesteros-Pérez [4] formula for correlation factor could only calculate up to four parallel paths, while
there were 420 possible parallel paths in this project from start to finish that must be merged. The original
formula used by Sculli and Shum [11] in their research were also applied to merge up to 14 paths only
and there were no detailed calculations provided in the research. In order to obtain results from both
PERT and M-PERT that can be compared, the next section of this research analyzed one of the
subnetworks in the project, which was the external works category.

Figure 5. Illustration of untangling paths into parallel paths

5
TICATE 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 508 (2019) 012034 doi:10.1088/1757-899X/508/1/012034

3.4. External Works Subnetwork Duration Mean and Variance of PERT and M-PERT
The External Works Subnetwork consisted of activities that were listed on Table 2. The activity mean
and variance for PERT and M-PERT were calculated the same way as Table 1.
Table 2. External works list of activities three point estimate, mean and variance
Act Optimistic Most Likely Pessimistic Act Mean PERT Var Correction Factor M-PERT Var
NO LIST OF ACTIVITIES Code O M P Te vTe K vTe
(week) (week) (week) (week) (week) (minggu)
4 STEEL STRUCTURE
4.5 MOTOR CANOPY E3 1 1 3 1.3333 0.1111 0.71429 0.07937
4.9 BILLBOARD E6 1 1 2 1.1667 0.0278 0.71429 0.01984
6 EXTERNAL WORKS
6.1 RIGID PAVEMENT H1 2 2 4 2.3333 0.1111 0.71429 0.07937
6.2 PAVING BLOCK PAVEMENT H2 2 2 4 2.3333 0.1111 0.71429 0.07937
6.3 ASPHALT PAVEMENT H3 1 1 2 1.1667 0.0278 0.71429 0.01984
6.4 CANSTEEN H4 1 1 2 1.1667 0.0278 0.71429 0.01984
6.5 PARKING LINE PAINTING H5 1 1 2 1.1667 0.0278 0.71429 0.01984
6.6 DRAINAGE SYSTEM H6 2 2 3 2.1667 0.0278 0.71429 0.01984
6.7 LANDSCAPING H7 1 1 2 1.1667 0.0278 0.71429 0.01984
6.8 SITE ENCLOSURE AND DIVISION H8 2 3 3 2.8333 0.0278 0.71429 0.01984
6.9 MISCELLANEOUS H9 1 2 3 2.0000 0.1111 1.28571 0.14286

The ADN for the external works subnetwork are given in Figure 6. After completely modelling the
network using forward pass and backward pass calculations, it was clear that the critical activities were
H2-H8-H6-H7-H9. The critical path produced a project duration mean and variance of 10.50 weeks and
0.3056 weeks respectively.

Figure 6. PERT Arrow Diagram Network of external works category

The PDN and merging procedure for M-PERT are shown through Figure 7. To sum it up, all of the
activities in the same path were first merged. After all the paths has been formed, the three paths were
sorted from top to bottom according to their total mean value. Path A and Path B were merged, and its
results (Path AB) would then be merged with Path C. In the end, the merging produced M-PERT mean
and variance value of 10.50 weeks and 0.2817 weeks respectively. It produced the same mean value as
PERT and lesser variance than PERT. The different variance value indicated that M-PERT would
provide a project duration with a more homogenous distribution to its mean than PERT. If these two
results were to be drawn as normal distribution curve, M-PERT’s normal curve would be leaner than
PERT’s curve as it has greater number of data with value closer to its mean.

6
TICATE 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 508 (2019) 012034 doi:10.1088/1757-899X/508/1/012034

(H6 + H7 + H8).
(H1 + (H4.H5)max + H3)max PATH AB H9 END
6,17 0,0595 2,00 0,14

START H2 E3 + E6 PATH C
2,33 0,08 2,50 0,0992

[{(H6 + H7 + H8).(H1 + (H4.H5)max + H3)max}.(E3+E6)]max

START H2 H9 END
2,33 0,08 6,17 0,0595 2,00 0,14

[{(H6 + H7 + H8).(H1 + (H4.H5)max + H3)max}.(E3+E6)]max + H2 + H9

START END
10,50 0,2817

Figure 7. Merging Procedure of External Works PDN for M-PERT

3.5. External Works Subnetwork Duration Mean and Variance: Monte Carlo Simulation
Monte Carlo simulation was performed using @RISK (trial version), an add-in software for Microsoft
Excel. The desired activity distribution is selected, then the three-point-estimate of activities were
entered into the software as input. Simulation results after 1,000,000 times of iteration are displayed in
Figure 8. The parameters in the right side of Figure 8 necessary to compare PERT and M-PERT results
with Monte Carlo simulation were the mean of 10.5002, standard deviation of 0.5301, minimum value
of 8.5364, and maximum value of 13.1692.

7
TICATE 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 508 (2019) 012034 doi:10.1088/1757-899X/508/1/012034

Figure 8. Distribution of external works duration produced by @RISK (trial version)

Comparison of the Monte Carlo Simulation, PERT and M-PERT results were created using Microsoft
Excel. The comparison was presented in the form of cumulative distribution from the three methods
(PERT, M-PERT and simulation) for X value (external works duration) of 8.50 to 13.00 weeks. As seen
in Figure 9, the results from PERT, M-PERT and Monte Carlo simulation were so close that the
cumulative distribution curve of the three methods were too difficult to be observed visually. Hence, a
new indicator was introduced to measure the difference between Monte Carlo simulation and PERT or
M-PERT; the Root Mean Square Error (RMSE) value.

Figure 9. Comparison of external works duration’s cumulative distribution from simulation, PERT, and
M-PERT

8
TICATE 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 508 (2019) 012034 doi:10.1088/1757-899X/508/1/012034

The RMSE calculation was done twice, as there were two comparisons observed; the simulation results
to PERT results; and the simulation results to M-PERT results. First, difference of probability obtained
from two methods were calculated for each X-value from 8.50 to 13.00. To calculate the RMSE value
of those two methods, the difference value must be squared and then summed. The total was divided by
number of observation and the square root value of that is the RMSE value. In Microsoft Excel, simply
list the cumulative probability for each observation (X-value) from the two methods as shown in Table
3. The cumulative probability value of PERT was computed by Excel formula of normal distribution
(“=NORM.DIST(…)”) using the mean, standard deviation and the desired X-value. The same also goes
for M-PERT. Difference is calculated by subtracting the “PERT” cell data from the “Simulation” cell
data. After difference for all observation has been calculated, enter the following formula at the bottom
cell as the RMSE formula: “=SQRT (SUMSQ (all DIFFERENCE cells)/COUNTA (all DIFFERENCE
cells))”. From Table 3, the RMSE value between simulation results and PERT results was 0.0053961.
Table 3. Sample for RMSE calculation between simulation and PERT
SIMULATION PERT DIFFERENCE
X (μ = 10,5002; σ = 0,5301) (μ = 10,50; σ = 0,55277) (SIMULATION - PERT)
8,5 0,0001 0,0001 -0,0001
8,6 0,0002 0,0003 -0,0001
8,7 0,0003 0,0006 -0,0002
8,8 0,0007 0,0011 -0,0004
8,9 0,0013 0,0019 -0,0006
9 0,0023 0,0033 -0,0010
. . . .
. . . .
. . . .
. . . .
13 1,0000 1,0000 0,0000
ROOT MEAN SQUARE ERROR (SIMULATION VS PERT) 0,0053961
The same steps were also performed for RMSE value between simulation and M-PERT. Results were
shown in Table 4. The RMSE value between simulation and M-PERT was 0.0001818. From here the
conclusion was clear that M-PERT was more accurate than PERT as it produced closer results to Monte
Carlo simulation than PERT.
Table 4. Sample for RMSE calculation between simulation and PERT
SIMULATION M-PERT SELISIH
X (μ = 10,5002; σ = 0,5301) (μ = 10,50; σ = 0,5308) (SIMULATION - M-PERT)
8,5 0,0001 0,0001 0,0000
8,6 0,0002 0,0002 0,0000
8,7 0,0003 0,0003 0,0000
8,8 0,0007 0,0007 0,0000
8,9 0,0013 0,0013 0,0000
9 0,0023 0,0024 0,0000
. . . .
. . . .
. . . .
. . . .
13 1,0000 1,0000 0,0000
ROOT MEAN SQUARE ERROR (SIMULATION VS M-PERT) 0,0001818

4. Conclusion
A research has been conducted to compare two probabilistic scheduling methods; PERT and M-PERT.
Using the external works as the analyzed project, M-PERT provided better estimation results than PERT

9
TICATE 2018 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 508 (2019) 012034 doi:10.1088/1757-899X/508/1/012034

as it produced the same value of mean duration with smaller value of variance. M-PERT also surpassed
PERT in terms of accuracy as it produced smaller RMSE value to the Monte Carlo simulation than
PERT. Although implementing M-PERT in a complex project was a challenge that has yet to be
answered, its principles can be used to develop a probabilistic scheduling software to assist scheduler’s
work in a construction site.

5. References
[1] Project Management Institute 2017 A Guide to the Project Management Body of Knowledge 6th
edition (Pennsylvania: Newtown Square)
[2] MacCrimmon K R and Ryavec C A 1964 An analytical study of the PERT assumptions
Oper. Res 12(1) 16–37
[3] Hajdu M and Bokor O 2014 The effects of different activity distributions on project
duration in PERT networks Procedia Soc. Behav.Science 119 766–75.
[4] Ballesteros-Pérez P 2017 M-PERT: Manual Project-Duration Estimation Technique for
Teaching Scheduling Basics J. Constr. Eng. Manage. 143(9) 1-12
[5] Cox M 1995 Simple normal approximation to the completion time distribution for a
PERT network Int. J. Project Manage. 13(4), 265–70
[6] Herrerías-Velasco J M Herrerías-Pleguezuelo R, Van Dorp J R 2011 Revisiting the PERT mean
and variance Eur. J. Oper.Res. 210(2) 448–51
[7] Lu M and Lam H C 2009 Transform schemes applied on non-finish-to-start logical relationships
in project network diagrams J. Constr. Eng. Manage. 135(9) 863–73
[8] Elmaghraby S E and Kamburowski J 1992 The analysis of activity networks under generalized
precedence relations (GPRs) Manage.Sci. 38(9) 1245–63
[9] Pritsker A A B 1966 RM-4973-NASA, (Santa Monica: Rand Corporation)
[10] Mehrotra K, Chai J and Pillutla S 1996 A study of approximating the moments of the job
completion time in PERT networks J. Oper.Manage. 14(3) 277–89.
[11] Sculli D and Shum Y W 1991 An approximate solution to the PERT problem Comput.
Math. Appl. 21(8) 1–7
[12] Pritchard C L 2001 Risk Management: Concepts and Guidance 3rd edition (Arlington:
ESI International)
[13] Van Slyke M R 1963 Monte Carlo methods and the PERT problem Oper. Res. 11
839-60.
[14] Wyrozębski P and Wyrozębska A 2013 Challenges of project planning in the probabilistic
approach using PERT, GERT and monte carlo Journal of Management and Marketing 1(1) 3

10