PERT

The PERT method of drawing and computing with network analysis was devised
to take account of the difficulty of estimating the durations of activities which cannot be
established conclusively from past experience. In the case of construction work,
excavation, for example, has been carried out under many varied conditions and in all
types of soil. The experience of conducting this operation is, therefore, very broad, and,
in estimating durations for carrying out future work, there is little doubt about the
estimated normal durations if the facts about the excavation conditions are known.
Where it is necessary to estimate the possible duration of such work as research or
development, or work that is being carried out for the first time, then the basis for
estimation is by no means as sound as that for construction works. It is for this type of
work that the PERT statistical approach was developed. Statisticians have relatively
simple means for the determination of uncertainty in quantitative terms and such
methods, though they have been challenged in this context, are the basis for the PERT
approach to network analysis, PERT methods do take account of the activity which is
likely to have wide range of durations. An activity in CPM may be estimated to have a
most likely duration of ten days. It may be that, due to the nature of the operations, the
possible range of the duration could be from two to eighteen days or, for a different
operation, it could be from eight to fifteen days. No distinction between these two types
of activity is made in CPM; the duration is stated to be 10 days and the probable
uncertainty is not considered at all.

Other than the different approach to estimating the durations of activities, PERT is
known as an event-orientated method. In the PERT method, certain events are selected
throughout the programme to act as milestones for the programme. These are events
which are important in the overall programme and against which progress in general can
be measured. These milestones are labelled with a specific title, for example,
Completion of all foundation work. With CPM, the tendency is to refer to the activity
and not to the event. When a number of events or milestones in the PERT network has
been established, these are linked together by arrows in the same way as the CPM
network. The expenditure of resource is assumed to occur on the activity arrow in PERT
and in CPM.

One commonly sees frequency distributions of various topics representing the results of
the surveys which have been made in specific fields. Such a frequency distribution is
reproduced in Fig. 9.13. The distribution represents the spread of times which a well-
defined activity has taken in the past. This might be something which is carried out
fairly frequently in a maintenance procedure or a factory workshop, for example. On the
vertical axis of the graph is shown the frequency occurrence for any duration and on the
horizontal scale the actual duration time as it was observed. From Fig. 9.13 it will be
seen that the activity took seventeen days for an observed number of times of 100. Other
numbers of occurrences for other durations are plotted. If the interval between the
activity duration is reduced and the number of plots will tend to a smooth curve, an
example of which has been superimposed upon the figure in 9.13. Such a frequency
distribution was assumed by the originator of PERT and in this case he assumed that the
plots of durations of the activities almost fitted a beta distribution curve, a curve the
characteristics of which are well known to statisticians. Having assumed that the beta
distribution was a satisfactory model for the estimation of duration time, it was then
possible to develop the equation
𝑎 + 4𝑚 + 𝑏
𝑡𝑒 =
6
Where : m = most likely duration of the activity
a= the optimistic time, that is the shortest time which could be anticipated for the
activity
b = the pessimistic estimate of the time, that is, the duration of the activity assuming that
everything goes at its worst.
Statistically it was assumed that the optimistic and pessimistic times would bc achieved
on only one occasion in every hundred. Figure 9.14 represents the situation of estimating
activity times in PERT related to the beta distribution.
If the most likely time for the duration of a certain activity is estimated to be ten days,
the optimistic duration is estimated to be three days, and the pessimistic duration of time
is taken to be fifteen days, then these three estimates of duration can be reduced to one
for the activity by substitution in the formula above
3 + (4 × 10) + 15
𝑡𝑒 = = 9.7 𝑑𝑎𝑦𝑠
6
It is usual in calculating durations for a PERT network to allow one decimal place. In the
above example the time to be shown against the activity would be 9.7 days-
The beta distribution curve is used in a further way in the PERT network to calculate a
variance for each activity. The variance depends upon the difference between the
pessimistic and optimistic times estimated. Referring to
Fig. 9.14, it will be seen that if the difference between a and b is large there is less
likelihood of meeting the duration estimate than if the difference between a and b is small.
Subsequently, a standard deviation may be calculated for the whole project represented
by the network and, by the use of statistical probability tables, a figure on the scale of 0—
1 can be calculated which will give a probability of achieving the time schedule of the
project as a whole and as computed by the use of the network. If, for example, a figure of
0.5 is obtained by the use of the probability tables, there would be what is commonly
known as a fifty-fifty chance of the target date being achieved.
Since the days when CPM and PERT were developed independently, there has been a
considerable overlapping between the two methods so that the identity of one is no
longer clear from the identity of the other. In construction work, with the wide variety of
experience which has been achieved, the estimator of duration times likes to think that
he can do this with considerable accuracy. The use of the three time estimate aspect of
PERT is not widely adopted for construction work.
Example 9. I
The arrow diagram in Fig. 9.15 represents the logical sequence of activities in a research
and development programme for the design and construction of a reinforced concrete
nuclear pressure vessel. Because of the uncertain nature of the programme it is decided to
analyse the arrow diagram using three estimates of duration for each activity. The three
duration estimates (in weeks) are shown against each arrow in the figure. Thus for arrow
1—2, the estimated optimistic, most likely and pessimistic durations are 6, 9 and 15 weeks
respectively.
 What is the probability that the programme will be completed in 33 weeks?
The first step in the analysis is to calculate the expected time, te, for each activity from
the general formula stated above. It should be noted that the mean of the three given
durations will only be equal to the value of te where the optimistic and the pessimistic
durations are symmetrically placed about the most likely duration. Activity 2—4 in Fig.
9.15 is an example. The values of te give a measure of the central tendency for the beta
distribution of the activity durations. In order to measure the spread or range of a
distribution, a standard deviation is used. A standard deviation is a measure of the
variability of a distribution, and readers should familiarize themselves with the statistical
terms and definitions used here. The standard deviation, s, of a beta distribution can be
found from the formula

(𝑏 − 𝑎)
𝑠=
6
Where a and b in this context are as previously defined. The result of calculating values
of the mean, 𝑡𝑒 and the standard deviation, s, for each activity in the network summarized
in Table 9.3.
We next make use of the central limit theorem which, in simple terms as applied to a
PERT network, states that:
Where a series of sequential independent activities lie on the critical path of a network the
sum of the individual activity durations will be distributed in approximately normal
fashion regardless of the way in which the individual activity durations are themselves
distributed. The mean of the distribution of the sum of the activity durations will be the
sum of the means of the individual activities and its variance will be the sum of the
variances.
The variance is the square of the standard deviation for a distribution. It is now possible
to calculate the standard deviation of the approximately normally distributed completion
time for the network of Fig. 9.15 as follows:

Standard deviation = √(1 2 + 1.33 2 + 1 2 + 0.5 2 + 0.5 2 + 0.5 2 ) = 2.13 weeks

The standard deviation has been found by calculating the square root of the sum of the
variances for the activities lying on the critical path through events
1-3-5-8-10—11. The length of the critical path is the sum of the means of those
activities forming part of it, and in this case totals 30.5 weeks.
By using the central limit theorem, it can now be stated that the completion time of the
project is distributed normally, having a mean of 30.5 weeks and a standard deviation of
2.13 weeks. The properties of a normal distribution are well established and set out in
statistical tables. Figure 9.16 illustrates the distribution of the end event times showing
also the position of the 33 week completion. The probability of completion in 33 weeks
is represented by the shaded area under the distribution curve as a proportion of the
whole area. In terms of standard deviations, the 33 week ordinate is 2.5/2.13 = 1.17 to
the right of the mean and (from tables) the shaded area is 0.8790 of the total. The
probability of completing the work in 33weeks is therefore 0.8790- a high probability.