Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
by
Jacques Hallak
The opinions expressed in this lecture are those of the author and do
not necessarily represent the views of the Institute.
i
IIEP/TM/37/69
CONTENTS
Pases
I. GENERAL PRINCIPLES 1
le Definition 1
2«, Programme analysis 2
3« The construction of the network 6
Conclusion 9
II. PERT-TIME 10
1. Estimating the duration of the activities
of a project 11
2 e Time calculation 13
3«. The concept of slack or float 16
Ц-* The critical path 18
5 » The probability of respecting a timetable 20
6e An example of application: research project
on education administration 23
(a) Preliminary over-all analysis 23
(b) Detailed analysis 25
Selected Bibliography 33
ii
XIEP/TM/37/69 ~ P a S e 1
I. GENERAL PRINCIPLES
This is borne out, for example, by the recent adoption of PERT for
major educational research and development programmes in the United States.
L Definition
2. Programme analysis
DIAGRAM No. 1
Objective of
the programme
Etc
Etc
Etc
ИЕР/ГМ/37/69 - Page 3
We would note here and now, and we can confirm it later, that once
the activities reach a certain number, a programme cannot be handled with
out the help of electronic machines. It is precisely owing to computers
that the field of application of PERT has developed in all spheres.
(i) What is the objective and what is the starting point towards
achieving it?
More precisely, the first thing is to find the initial event (or
events). It is easy to identify. It is the one which has no predecessor
events. The next thing is to look for all the activities subsequent to
the initial event. Each of these activities necessarily leads to an event.
The question then is, what is the starting event from which that activity
can be undertaken? We thus get a table of events completed by immediately
preceding or following activities. The table is checked (and, if necessary,
completed) by examining the activities preceding each event.
(l) The event or events immediately following another event with no inter
mediate event or events, are called Successor events?. The event or
events immediately preceding another event with no intermediate event
or events are called fpredecessorч events.
ИЕР/ГМ/37/69 - Pag© 5
Events Activities
The following matrix shox^s the final result of the analysis of the
sub-project. It indicates the logical sequence of events and activities.
D С A В E F
D b
С d a
A с
В e
E f
F Objective
DIAGRAM No, 2
D
b
Ш
J^.
DIAGRAM No. 3
- that all events and all activities are clearly shown on the graph.
In the first place, the paths plotted on the graph must be followed
from the origin to the objective to check the logic of the sequence adopted.
DIAGRAM No. 4
DIAGRAM No, 5
,0 ,© ,0 0) *0- - -
.0^ ^0 *©
O I 2 3 4 5 Months
The check of the paths may also show that some of them do not lead
to the objective event. This means;
DIAGRAM No, 6
0-0
•0-0-0-0-0 -> Objective
Conclusion
Those are the broad outlines of the general principles of the PERT
method., From these principles., it clearly follows that PERT consists
essentially of analysing logically and in detail all the events and activi-
ties which go towards the execution of a project in order to programme
them with optimum effect«, This is enough to introduce the potential uses
of PERT in educational planning. It is quite certain., for example that
school building projects may constitute a preferential field of application
for PERT«, Another field is the execution of applied research or experimen-
tation projects. Similarly., the preparation of education plans., owing to
its increasing complexity and the fact that it calls upon a considerable
number of participating agencies^ may be built around planning based on
PERT, Finally} the application of techniques allied to PERT to the adminis-
tration and management of higher education establishments seems also destined
to extend and to become progressively more firmly established«,
II. PERD-TIME
After the project has been analysed and the network or graph has
been constructed (see Part l) it is valuable to complete this network by
indicating the duration of the various activities identified in the project.
In practice, as has been shown, while events consume neither time nor money,
activities, on the other hand (other than dummy activities) consume time«
Most likely estimate; this is the most likely time for the com
pletion of the activity. It is the time which would have been
given in any event if the executants had been asked to give a single
estimate instead of three.
formulai
te = a + 4m 4- b (l)
DIAGRAM No. 7
(l) If a = 9, b = 19 and m =
IÏEP/DM/37/69 - page 13
2. Time calculation
Once the standard times have been estimated for all the activities
of the project, they are plotted on the network above the activity arrows,
as shown in the following diagrams
(a) Early start date (or fearliest expected time')? once the times
are shown for each operation on the graph, it is possible, by following the
paths which make it up, to calculate the dates at which the various events
of the project can be achieved. If there is more than one path leading to
an event, the event is obviously not achieved until all the activities pre-
ceding it have been completed. The date which corresponds to the time of
the activities on all paths is the !earliest expected time* of the event.
For example, in the above diagram, the earliest expected date of event k is
9 (i.e. 2 + 7 ) and not 7 (i.e. k + 3).
Thus, the ?early start date' or 'earliest expected time1 for reaching
an event is obtained by taking the maximum time for the activities on all
the paths leading to the event.
Exercise
Te=0
T°l = T2 + 1 = 1
T22 = T2 + б = б
T23 = T2 + 12 = 12
T24 r= Maximum of (T21 + 13 or T22 + 11) = 17
T25 = T23 + 10 = 22
T25I = Maximum of (T24 + 9 or T25 + 7) = 29„
11ЕРД,М/37/б9 - page 15
(b) Late start date (or 9latest allowable time*)J in general, the
question of time and completion dates arises differently. A time limit for
completing the project is decided upon, and it then remains to fix the
f
latest allowable times* for the successive events of the project«, In other
words, the completion date for a project being specified by contract, what
is the latest date for reaching each event? Those responsible for the dif
ferent activities are all concerned by this question since they want to know
the time limit within which they must complete the activities for which they
are responsible«, To arrive at the latest allowable time., or late start
dates for each event, it is necessary to start from the objective date and
work back to the originj by subtracting from the duration of each preceding
activity its latest allowable date, we arrive at the latest allowable oíate
of the preceding event. If, however, more than one value is obtained for
the latest allowable time of an event, the lowest is obviously chosen* For
example, if two activities start from event SA? the latest date for the
start of fACs will be T ? c - dc and the latest date for the start of fAB?
will be T'B - dB, The latest date at which event 9A* must be completed is
therefore the earlier of these two dates.(l)
0
It is thus possible, step by step, to arrive at the latest dates
for all the events of the project.
Exercise
Taking the network given above, assume that the directed date for
5
251* is 50. In order words, T f 251 = 30, Calculate the latest dates for
event 2„
(l) The symbol T is generally used for the earliest dates and the symbol T
for the latest dates.
TIEP/ÏM/37/69 - P a g e 1б
251 30- 30
25 30-7 23
?A 30-9 21
21 21-13 8
22 21-11 10
23 23-10 13
2 8-1 1
10-6
13-12
Once the latest date and the earliest date for each event have been
assessed., sufficient information is available to calculate the slack or
?
floatf available for achieving each event. If the earliest date and the
latest date coincide, there is no margins the slack is nil, and any delay
in reaching the event will have a repercussion on following activities.
If, on the other hand, the latest date is subsequent to the earliest
date for an event, the difference between these two dates constitutes a
margin of manoeuvre for the project manager, since he has a certain latitude
in carrying out the activity in question or supervising the event. The value
of the slack T* - T is then positive, and measures the magnitude of the
available time margin.
This concept of slack applied to events, may also apply to the activi
ties of a programme. There may thus be a slack between the earliest starting
date and the latest starting date of an activity. Similarly, there may be
slack betitfeen the earliest finishing date and the latest finishing date.
The simultaneous intervention of different types of slack very soon makes
the network of a PERT project very complicated* Without wishing to go too
deeply into this idea of slack, I nevertheless think it useful to show where
its application may lead in two important special cases.
IIEP/ÏM/37/69 - page 17
T1 = 2 T =7
!
©— ^
T = 1
2
Since the earliest date of event *2S is 1, and of event *1* is 2, and the
time of the two activities 1-3 and 2-3 is successively 4 and 6S the earliest
date for event 9 3 f is 7» But it is clear that if activity 1-3 starts at the
earliest date of event sl', which is also the earliest date for starting
this activity, then the earliest finishing date for activity 1-3 does not
coincide with the earliest date of event *3?e There is a float of 7-6 « lj
this is called the ffree float* of activity 1-3.
Since the latest dates of events 'З1 and f 4 f and successively 8 and 9*
and the duration of activities 1-3 and 1-4 4 and 7* the latest date of event
f
l* is obviously 2. But it is clear that, even if activity 1-3 does not
start at the latest date of event ? 1* but two time units later, there will
be no change in the rest of the programme. There is therefore a slack
between the latest finishing date for event f l f and the latest starting
date of activity 1-3« This kind of slack is called findependent'. Here
again,, resources can be economized by eliminating the independent slack,
thus making the achievement of the project more efficient»
On the other hand, a nil slack means that the times are strictly
observed (adequate resources) and that the earliest and latest dates of the
event coincide. The event is then said to be critical. Any delay in
reaching a critical event has the effect of delaying all subsequent activi
ties, and may therefore imperil the completion of the project within the
time fixed. It is quite clear that if the value of the slack of an event
is negative, the event is also critical.
Exercise
Calculate the latest times and the slack of the events on the
following network. Indicate the critical path of the programme»
T„=8
Тг=31
Events T» T Slack
6 31 31 0
5 27 27 0
4 15 8 7
7 20 8 12
3 16 16 0
2 7 7 0
1 0 0 0
У 2
T,
"E
where;
t is the timetable date in question
s
- T is the earliest date of the objective event
e
the denominator of the fraction measures the standard varia
tion, i.e. the dispersion of the variable which equals the
square root of the sum of the variances of all the activities
preceding the objective event. It will be remembered that
the variance of an activity is given by the formulas
Activities a П1 b Activities a m b
1-2 7 10 13 2 » 5 10 12 17
1 - 3 16 18 20 3 - 5 20 24 34
1 - 4 7 8 9 5 - 6 2 3 7
4 - 6 14 18 28
Assume that the data are in dayse The agreed time for reaching
event 5 is 45 days from the start of the work. Calculate the
probability of respecting this date. The average duration of
the 7 activities obtained by the equation te = (a + 4m + b)/6
are as follows:
1-2 ; 10; 1-3 % 18; 1-4 : 8; 2-5 : 12.5l 3-5 % 25; 4-6 % 19;
5-6 : 3.5
The earliest dates for the different events are therefore:
T_ = 0 Т л = 10 T , = 18 T,. = 8
1 d 3 4
T = Max /"10 + 12,5 or 18 + 25J 7 = 43
T 6 = Max /~43 + 3.5 or 8 + 19_7 = 46 „5
The numerator of Z is therefore 45-43 = 2
The critical path leading to event 5 is:
1 - 3 and 3 - 5
The variances of these activities are given by the equation:
¿a
*' -fcy - С -(*T - 116
» -(«У
The sum of the variances is therefore —=7*
and the standard variation of the variable is: V •==•?» orXZTl»8
ИЕР/ГМ/37/69 - page 22
The table shows that the probability of observing the time limit
of 45 days is 0.8643 or more than 86 per cent.
NORMAL DISTRIBUTION
(Laplace-Gauss Law)
z PR z PR
3.0 0,9987
IIEP/FM/37/69 - P a § e 23
Four important predecessor events (M., M_5 M_ and 1VL) govern the
achievement of these sub-objectives: 'Missions effected in countries A,
В, С and D*.
Finally5 the initial event OR: 'field and limits of the project
established'„
DIAGRAM No, 8
N/
ч/-
M/-
MC
v/
D
v
IIEP/FM/J7/69 - page 25
In the next phase the intermediate events between the main events
must be analysed in greater detail;
Between OR and S
f
Events Note of proposals to aid agencies prepared* 1.
f
Note of proposals despatched' 2.
'Contracts negotiated8 5»
'Contracts signed' 4„
f
Activities Prepare a note of proposals' OR-1
Between S and R
'Bibliographic note* 6»
'Meeting started1 r
о
'Bibliography discussed1 r
*Studies discussed* r
'Notes discussed* r_
'Discuss notes' r -R
22,
'Select A, Bs C, D* 21-22
8
Ask for agreement' 22-23 (A, B, C, D)
OB,
The PERT network of the project can be constructed from the elements
of the detailed analysis. In the following table the standard times of
the activities of the project have been expressed in weeks (one year =
45 weeks).
Standard times
Activity te Activity te
OR-1 4 13-19 0.5
1-2 4 19-20 1
2-3 1 20-21 1
3-4 1 21-22 4
4-S 0.5 22-23 0,2
OR-5 2 23-24 1
5-S 2 16-24 0
S-6 1 24-25 1
S-7 4 25-26 0,2
7-8 1 26-27 2
S-9 0.5 27-28 2
S-10 1 28-М.1 0,2
10-r 1 M-29 4
о
7-11 1.5 29-30 2
11-12 1 ЗО-З1 2
Г +Гт 0,2 З1-З2 1x4
0 1
0.3 32-33 0,2
г -г
1 2 0o5 33-A s B 5 C 5 D 4
Iu-Tp-R 1 ЗО-34 1
15-16
R-13 0,5 34-35 3
R~l4 1 35-36 6
14-15 p 36-37 2
R~17 0e2 37-З8 3
17-18 4 38-OB 12
И Е Р / Г М / 3 7 / 6 9 - page 32
0 R - l - 2 - 5 - ^ - S - 7 - l l - 1 2 - r - r 1 - r 2 - R - 1 7 -
18 - 20 - 21 - 22 - 23 - 24 - 25 - 26 - 27 - 28 - M - 29 - 30 -
34 , 35 . 36 - 37 - 38 - oB
Activity a b Activity a b
0R-1 1 7 21-22 3 5
1-2 2 6 22-23 0.2 0.2
2-3 0.5 1-5 24-25 1 1
3-4 0,5 1.5 25-26 0.1 0.3
4-s 0„2 oe8 26-27 1.5 2.5
S-7 3*5 4.5 27-28 1.8 2.2
7-11 0,5 4.5 28-М 0.2 0,2
11-12 0.5 1.5 23-24 0.8 1.2
r -r. 0,2 0,2
0 1
r —r 0*3 0,3
1 2
r2-R 0,5 0,5
R-17 0*2 0,2
17=18 3 3
20-21 0*8 1.2
IIEP/TM/37/69 - page 33