Sei sulla pagina 1di 63

Chapter 11:

PERT, CPM and Other


Networks
Major Difference And Similarities Between PERT
And CPM
a. Program Evaluation Review Technique (PERT)
1. Probabilistic tool
2. Used to plan and control time
b. Critical Path Method (CPM)
1. Deterministic tool
2. Used to control both time and cost of project
Similarities:
Allow users…
1. To manage other resources
2. To trade off resources
3. To analyze different types of schedules
4. To balance the use of resources
Purpose of PERT And CPM:
1. To objectively identify critical activities
2. Can also tell us how close the remaining activities are to becoming
critical
Information Provided to Managers:
1. Which activities are critical
2. Which activities are noncritical
3. The amount of slack on each noncritical activity
Advantages of PERT and CPM:
a. Detailed Planning
b. Commitments and Communications
c. Efficient Monitoring and Control
d. Identifying Potential Problem Areas
e. Proper Use of Resources
f. Rescheduling
g. Government Contracts
h. Easily Understood
h. Adaptable to Computers
j. Tools for Decision Making
k. Assess Probability of Completion (in PERT only)
l. Cost-Time Trade-offs (in CPM only)
The PERT And CPM Procedure:
=Formulation
STEP 1: Analysis of the project.
STEP 2: Sequence the activities.
STEP 3: Estimate activity time and costs.
=Solution
STEP 4: Construct the network.
STEP 5: Event analysis
STEP 6: Activity analysis
=Analysis & Application
STEP 7: Monitoring and control
STEP 8: Resource utilization
Formulation:
1. Analysis of the project- list of activities is agreed on and is
clearly defined and responsibility is assigned to the proper
department heads.
2. Sequence the activities- sequence of execution is determined
3. Estimate activity times and costs- determining the required
duration for each activity, designated as expected time.
CPM – activity duration is considered to be deterministic
PERT – the duration is unknown so we use the estimates
Solution:
4. Construct the network – PERT/CPM can be solved
manually or through computers
Two basic network approaches:
a. Event oriented
b. Activity oriented
Dummy Activities
In the construction of a network, care must be taken to
assure that the activities and events are in proper sequence. One
device that helps proper sequencing is dummy activities. They
are characterized by their use of use of zero time and zero
resources; their only function is to designate a precedence
relationship.
Network Without Events (Activity-Oriented)
It is used mainly in computer analysis
where events are not used at all. Figure 11.8
shows a network composed of activities only (a)
and its equivalent event-oriented network
Time-Scaled Networks
One in which the activities and events and
are located by a time scale along the horizontal
axis. The estimated duration of the activity is
shown as a bar whose length is proportional to
the duration. PERT can be viewed as an
extension of a simple Gantt bar/chart.
5. Event analysis – used primarily in PERT because it is especially
convenient for risk analysis involving probabilities of completion.
The ff. procedure is used in event-analysis:
a. Enter time estimates on the network – once network is
completed, activity durations
(te) are entered on the diagram
b. Compute the earliest and latest dates for all events –
based on 2 important concepts
=Earliest possible event date – Te (ET) – for an event to
occur is immediately after all the preceding activities have been
completed.
=Latest allowable event date – Tl (LT) – is the latest date the
event can occur without causing a delay in the already-determined
projects completion date. The completion date for the project can be
either the earliest possible completion date or any other agreed-on
date, unless otherwise stated, we will use the earliest possible
completion date for our calculations.
Step 1: conduct a forward pass: find Te for each event. The
duration of each path leading to the event is computed. If several
paths lead to an event, then the path with the largest elapsed time
is selected. Computed starting from the left.
Step 2: conduct a backward pass: find Tl for each event. To
compute each Tl, start from the last event and work backward to
the first event. Or compute starting from the right side of the
network.
c. Find the slack on the events and identify critical
events – the difference between Tl and Te for each event is
defined as slack.
S= Tl – Te
2 cases of Slack event:
1. When Tl = Te for the last event. In this case, slacks in the
network can either be zero (0), whereupon the events are
called critical events, or larger than zero (0), whereupon the
events are considered to have positive slack.
2. When Tl ≠ Te for the last event. In this case, the critical
events are defined as those events with the minimum slack,
which can be negative.
d. Find the slack on the activities and identify critical
activities – this slack tells us how long the activity can linger
without delaying the entire project.
Activity slack = TL – Te – te
SLACK – tells how long the event can linger without
delaying the entire project. Also called total float (TF).
2 Cases of Slack Activities:
1. When Tl = Te for the last event. An activity with zero (0)
slack is defined as a critical activity.
2. When Tl ≠ Te for the last event. The activities with a
minimum slack are the critical ones.
e. Find the critical path – the critical path is the path in
the network, leading from the beginning of the project to its
end, all of whose activities and events are critical. This
definition implies that if Tl = Te for the last event, then there is
zero (0) slack on the critical path. Otherwise, the critical path
is that path with the minimum slack on it.
Additional characteristics of critical path:
1. There can be more than one critical path in the network.
2. The critical path is the longest (time wise) path in the network.
6. Activity Analysis
Let:
ES = Earliest start time for an activity. This time is
equivalent to the Teof the event from which the activity starts.
EF = Earliest finish time for an activity. Assuming the
activity started its ES and lasted its planned duration.
EF = ES + Te
LF = latest finish time for an activity. This is the latest
time by which an activity can be completed without delaying the
project. It is equal to the Tlof the event at the end of the activity.
LS = Latest start for an activity. This is the latest an
activity can start without jeopardizing the project’s deadline.
LS = LF – Te
Computing ES and EF (Forward Pass)
The procedure starts from the left and moves to the right.
Let the earliest start for the project be zero. (Alternatively, an
actual date can be used.)
Rule for computing the ES: the earliest time for an activity is
equal to the largest value of the earliest finish (EF) for all
activities ending at the event from which the activity starts.
Computing LF and LS (Backward Pass)
Once all ESs and EFs are computed, we start by setting
the LF of all final activities to the largest EF. Alternatively, a
desired finish date (larger than the largest EF) may be used as the
starting LF. Then, we sequentially compute the resulting LSs,
then the resulting LFs, and so on until the beginning node is
reached.
Rule for computing the LF: the latest finish time for an
activity entering a particular event is equal to the smallest value
of the LSs for all activities starting from that event.
Computation of the Slack
The regular slack, also called total float (TF), is computed
as: TF = LS – ES
Some additional Characteristics:
Regular Slack (Total Float) – denotes the maximum
slack variable, some of which may be shared.
Shared Slack (Slack on a Noncritical Path) – two or
more noncritical activities or noncritical events connected in a
series.
Free Float (FF) –a slack that represents the time any
activity can be delayed before it delays the earliest start time of
any activity immediately following. If there are several
noncritical activities in a series, only the last one will have a free
float.
Case when Tl≠Tefor the Project
If Tlis larger than Te for the last event, then the slack for
the last event will be positive.
If Tlis smaller than Tefor the last event, a negative slack
will result, indicating that the desired date cannot be achieved and
a delay of the magnitude of the negative value is expected.

A Critical Path Leading To an Event – the critical path


procedure outlined previously can also be used to find the critical
path leading to any desired event.

Use of Complete Enumeration


Critical path may be identified by listing all positive paths
leading from the beginning of the project to its end. The path with
the longest path is longer than any other, its completion gives
enough elapsed time to complete all other paths. This guarantees
that every single activity in the network will be accomplished.
Analysis & Application:
7. Monitoring and Control
If the slack has a negative value, it is labeled negative slack. A
negative value for slack means that the project is behind schedule
8. Resource Utilization
The regular PERT/CPM analysis is limited to planning the
elapsed time. Because the planning is done prior to actual project
execution, it is not always possible to know the precise resource
availability, so we assume that there are sufficient resources for
executing the activities as planned.
Finding the best schedule – A trial-and-error
approach
1. Resource Leveling
An attempt is made to reduce peak resource
requirements and to smooth the weekly requirements within a
constraint on project duration
2. Resource Allocation
For a desired resource level, find the shortest
possible project schedule
Part B: Extensions
11.8 Estimating Activity Times in PERT
Three Estimates of Duration Time:
1. Optimistic Estimate (to): An estimate of the shortest
possible tune (duration) which the activity can be
accomplished. The probability that the activity will take less
than this is 0.01.
2. Most Likely Estimate (tm): The duration that would occur
most often if the activity were repeated under exactly the
same conditions many times. Equivalently, it is the time that
would be estimated most often by experts.
3. Pessimistic Estimate (tp): The longest time that the activity
could take “when everything goes wrong.” The probability
that the activity will exceed this duration is 0.01.
Computing the Weighted Average
Once the three estimates are obtained, their weighted average
is computed. This average, which is called the mean time of an
activity, te, is a weighted average of the three time estimates.
Three time estimates is computed using Equation
11.6:

The formula gives four times more weight to the most likely
estimate than to the pessimistic or optimistic estimates. The
division by 6, the sum of the weights, is to obtain a weighted
average.
Example: from table 11.5
11.9 Finding the Probabilities of Completion in PERT (Risk
Analysis)
The Consideration of Risk
PERT has more capabilities than just as a planning and
control tool. It can also be used to give management an indication
of risk in terms of project completion. This is a crucial analysis
that considers the chance of completing the project on, before, or
after scheduled dates.
Approaching the Normal Distribution:
The justification for the use of the normal distribution is
based on the central limit theorem: the sum of n independent
variables tends to be normally distributed as n approaches infinity.

Only 50% Chance of Completion by Earliest Time:


The project estimated completion time, TS, is computed as
the expected time, TE , for the last event. Therefore, there is a 50%
chance that the entire project will be completed by its earliest
projected time (27 weeks in our example from Part A0. However, a
“50% chance” is usually too low a confidence level for managerial
planning.
To calculate the probabilities, it is necessary to find the
standard deviations for the activities. The standard deviation of the
Beta Distribution of activity durations is given by Equation 11.7.
Assuming that the durations of the activities are independent
of each other, the variance of a group of activities (designated by V)
can be computed by adding the variances of the activities in that
group. The value of V is then expressed by Equation 11.9.
The Variance along the Critical Path:
Of special interest are the activities that comprise the critical path.
Example in Moose Lake project:

A Critical Path to each event:


The value of V can be computed, in a similar manner, for any event in
the network by considering the group of activities along the critical path
leading to the event.
The method described here is valid only if the following three
assumptions hold: (1) there is a large number of activities (at least 25) on the
critical path; (2) the activities’ completion times are independent of each other;
and (3) the noncritical paths are not relevant. If the above assumptions are not
valid, simulation must be used for the risk analysis.
Managerial Applications
The managerial questions raised at the beginning of this section—the
chance of completing the project in a certain desired time and the duration
necessary for obtaining any desired probability of completion—can now be
answered.
Figure 11.23 Chance of completing the project in 25, 27 and
30 weeks.

If the normal distribution table included negative numbers, the


probability could be read directly from them. Because this is usually not the
case, the probability for Z = +93 is read first (which is .8238). This value is then
subtracted from 1.0 (the total area under the curve), that is, 1-.8238 = .1762.
Thus, there is only a 17.62% chance of completing the project in 25 weeks or
less.
If Z is negative, the corresponding probability is always
less than 50%; and if Z is 0, the corresponding probability is
exactly 50%.
Goal-seeking Analysis:
The Variance of a Noncritical Path:
However, that there is a noncritical path whose variance
V is larger than, or even similar in magnitude to, the variance of
the critical path. What might its effect be on the probability of
completion? If equation 11.10 is used for new path, then the
probability of completion by the desired time might very well be
lower than that computed using the critical path. Therefore, if
more than one critical path exists, then V should be computed for
all such paths. To be conservative, use the path with the largest V
to compute probabilities of completion for dates after the
expected completion time, use the smallest V for probabilities of
completion for dates before the expected completion time.
11.10 The Critical Path Method (CPM): Cost-
Time Relationships
CPM analysis is used to evaluate various
alternatives of executing projects in those cases where it
is possible to expedite the execution of some or all of
the project’s activities. Expediting activities requires
additional resources, which means increasing the cost of
the project. However, considerable savings may be
realized in projects finished ahead of schedule.
Thus, the decision of how much to expedite may
be of great importance to management. The tool that
enables such an analysis is CPM.
Figure 11.24 CPM cost-time trade-offs for an activity:
The CPM Analysis:
The CPM analysis examines the total cost involved in
executing the project at various scheduled times, starting with either
the lowest cost-longest duration alternative or with the higher cost-
shortest duration alternative. The additional cost of expediting the
project can be compared with the possible savings from the
expedited completion.
The CPM analysis starts by solving the problem twice. First,
attention is paid only to normal times. Using the procedure outlined
in Part A of this chapter and assuming that the normal rime are the
te’s, a solution is derived, and its cost is also computed. Second, by
considering only the crash times as te;s, another solution is derived,
and its cost is also computed.
Once the two solutions are computed, the cost-time trade-
offs are used to find the least-cost plan for any number of weeks
between the all crash and normal plans. This cost can be compared
with the anticipated benefits.
Example:
A network of activities for a maintenance project is shown
in Figure11.26. The problem is to find the least-cost plan for
various project durations.
The normal time in days and cost, as well as the crash
time and cost, are shown in Table 11.6. The column “Cost slope”
indicates the incremental increase in cost when the duration of the
project is decreased by one day, computer from the Equation
11.12. Example for activity D:
Solution:
The first observation that can be made from Table 11.6 is
that if all activities are performed in the normal duration, the total
cost will be $1860. Second, if all activities are performed on a
crash basis, the total cost will be $2860.
All-Normal Basis: least-cost, longest time
Considering first all-normal times, the critical path can be
computed using the procedure shown in Part A of the chapter.
The results are shown in Figure 11.27. The critical path is 0-1-2-
4-5 for duration of 25 days and a cost of $1860.
All-crash Solution: most-expensive, shortest time
In a similar manner, the critical path of all crash duration
is computed Figure 11.28. The critical path is 0-1-3-4-5 for
duration of 17 days and a cost of $2860. At this stage, the
following analysis is performed:
a. Determine the minimum cost for the crash time of 17 days.
It was found it is possible to perform the project in 17 days. The
question is whether it is possible to perform the project in 17 days
but at a lower cost. To achieve a cost-reduction, the noncritical
activities could be performed at slower pace. This is called
expanding the activities.
Step 1: All noncritical activities are listed with their appropriate
cost slopes.
Step 2: The activity with the largest slope is selected first. The
largest savings can be made if this activity is expanded first. It
would be desirable to expand it as much as possible to achieve as
large a cost reduction as possible.
Step 3: the second largest cost reduction potential is
expanded next.
b. Determine the least-cost plan for any desired number
of days between all-normal to all-crash.
The normal schedule is the longest (slowest)
schedule for carrying out a project and costs the least.
On the other hand, the all-crash schedule is the fastest,
but is also the most expensive. In certain cases,
management needs to know he cost of carrying out the
project at some point between the fastest and he slowest.
Such a situation may develop, for example, when a
customer offers to pay a certain amount as a bonus for
finishing ahead of schedule.
11.11 Other Network Techniques
Project-Oriented Techniques:
PERT/Cost. The PERT technique is a time-
oriented method that helps plan and control a project’s
duration. PERT/Cost is an extension that permits the
planning and control of both time and cost. The basic
concept of PERT/Cost is that costs are to be measured
and controlled primarily on a project basis rather than
say on a departmental basis. Thus, individual activities
form cost centers for both accounting purposes and
managerial control. This is in contrast to conventional
cost methods, where organizational units such as
departments are the cost centers.
Other Techniques. Several other planning
techniques, including PERT II, PERT III, PERT IV, LESS,
TOPS, COMET, and PROPT (all extensions or
modifications of PERT or CPM), are available.
Graphical Evaluation and Review Technique
(GERT). The application of PERT assumes that all
activities must be completed before an event can be realized,
that events cannot be repeated, that all activities in the
network must be completed, that estimates follow the Beta
distribution, and that the critical path is the one with the
longest elapsed time, even though variances from those
mean time exist. GERT is an extension of PERT, where all
of the above assumptions are relaxed; that is, they are not
imposed any longer.
Nonproject-Oriented Network Techniques:
A network is a collection of nodes connected by
branches. Figure 11.35 depicts a general network. Other
examples are decision tree and PERT (event-oriented)
diagram.
11.12 Minimal Spanning Tree
Example:
The Northwood Telephone Company is planning a
network of telephones for a small, high-class shopping
center.
The lines to every telephone must be buried to
maintain the posh image of the center. The telephone
company wishes, of course, to minimize the necessary
digging, but due to existing electrical and utility lines, it
is not possible to dig just anywhere. Figure 11.36 shows
the feasible underground lines as arcs with their distances
shown. The problem is to determine which of these lines
to dig such that the total distance is minimized, yet every
phone is connected to the network.
In this case, the objective is to select a set of
branches in a network, from a feasible larger set of
branches, that will span (connect) all the nodes of the
network, yet minimize the sum of the branch lengths
(which equate to cost). This problem is called the
spanning tree problem. In solving this problem, we are
searching for the minimal spanning tree.
Solution Using the Greedy Algorithm:
A popular solution method is commonly referred as to the
greedy algorithm, because simply joining to the closest node at
each iteration of the solution turns out to be the optimal
procedure.
First, arbitrarily select any node and connect it to the
nearest node. We will start with 1 and connect it to 3, the closest
node. From the connected nodes 1 and 3, we search for the
shortest distance to any unconnected node—3 to 2 in our
example—and connect the two. We repeat this process, always
considering all connected nodes when searching for a new node
to connect, until all nodes are connected. In our example:
From 2 to 4
From 4 to 6
From 4 to 8
From 8 to 7
From 7 to 5
The optimal solution has a total distance of 178 based on Figure
11.37.

Notes:
1. There may be more than one optimal solution.
2. The total number of connecting segments is always n – 1,
where n is the number of nodes.
3. A more complicated problem is one which there is limits on
the capacities that can be transmitted along the arcs (a
capacitated spanning tree).
11.13 Shortest Route (Path)
The shortest route problem can be stated as follows: Find
the shortest route (in terms of distance, time or money) from a
given node in a network to any (or all) of the nodes in the
network.
Example:
A bank has six branches and a headquarters in the great
Los Angeles metropolitan area. The distances along the streets
are shown on the map. Some routes, such as from 1 directly to 5,
are not possible because of an intervening airport, railroad, or
park. Note that the arcs are not directed; that is, they permit travel
in either direction. Bank employees frequently travel from
headquarters 1 to the branches. The problem is to find the
shortest distance from headquarters to each branch.
Solution:
The algorithm used iterative in nature. In each iteration,
the shortest distance to one node is determined. Therefore, the
optimal solution will be reached in n – 1 iterations, where n = the
number of nodes.
Step 1: Determine the distance from headquarters 1 to
every node that can be reached directly. Label each such
destination node according to the following code:
Figure 11.38 Bank Route Problem:
Step 2: Box the initial node, then box the closest node to
the initial node.
Figure 11.39 The Optimal Solution:
Step 3: We now have three types of nodes:
a. Those closest that are boxed (1 and 3).
b. Those that are labeled (2, 3 and 4).
c. Those that are not labeled (5, 6 and 7).
Start from the latest boxed node, 3, and check the distance to all
unlabeled direct nodes. Because 5 and 6 are unlabeled, label them using the
distance from the initial node 1 through node 3.
For 5: The label is (3, 5) – 21.
For 6: The label is (3, 6) – 36.
Step 4: Check the labeled nodes, 2 and 4, for possible
improvements.
For node 2: Two alternatives exist: 1-2 for a distance of 20, or
1-3-2 for a distance of 25. Because no improvement can be achieved by
going through 3, we include that 1-2 is the shortest way. Therefore 2 is
boxed.
For node 4: The previous label indicated a distance of 25. However,
following route 1-3-4 results in a shorter path of 10 + 12 = 22. Therefore,
the node is relabeled (3, 4) – 22.
Figure 11.40 Improved Solution:

Step 5: Step 3 is repeated for node 2. At this iteration, the


label of 5 is checked and stays unchanged. Node 4 is boxed,
because no improvement is possible. Also, node 7 is labeled.
Step 6: Step 4 is repeated for node 4.
Figure 11.41 Second Improved Solution:

Step 7: Step 4 is repeated for node 5. No improvement is


possible, so the node is boxed.
Step 8: Step 4 is repeated for node 6. Improvement is
possible, so we box it too.
Figure 11.42 Optimal Solution:
The optimal solution is them read from the labels. From 1, the
shortest paths are:
11.14 Maximum Flow
The objective in this technique is to find the maximum amount of
flow of fluid, traffic, information, and so on, that can be transported through a
capacitated (capacity limited) network. Such a network is composed of a
source and a sink and connecting arcs and nodes. Flow in each individual arc
is measured by the amount that can be transported during a unit of time. Flow
can only be one-way along an arc, as shown by the arrow connecting points 1
and 2 in Figure 11.43. However, it is also possible to have two direct arcs
between nodes (e.g., points 2 and 3) so that flow can be two ways.
Figure 11.43 a capacitated network:
Solution Procedure:
Step 1: Trace a continuous path from the source to the
sink; for example, 1-2-3-4. (Do not consider any path where there
are one or more arcs with a zero-flow capacity)
Step 2: In this continuous path, determine the arc with the
minimum flow capacity. In our example, it is arc 2-2 with a flow
of 4; thus, the most we can transport through path 1-2-3-4 is 4.
Step 3: Reduce all the quantities along this path by 4 (the
maximum amount transported).
Figure 11.44 Network after iteration 1:
Iteration 2:
Repeat steps 1 through 3. This time, the path 1-3-4 is
considered. (Note that the order in which paths are considered
does not matter.) The minimum quantity is 6.
Figure 11.45 Network after iteration 2:
Iteration 3:
Repeat steps 1 through 3. This time, path 1-2-3-4 is
considered. The minimum flow is 3.
Step 4: Because all the paths have at least one arc with a
zero capacity, we stop. The maximum flow in the entire network
is 4 + 6 + 3 = 13.
Note: In some cases, it may be desirable to reverse a previously
assigned flow through an arc to increase the total source-to-sink
flow. This is acceptable.
Thank You For Listening! 
Cruz, Pauline Joyce
Suldan, Judy Ann
Valencia, Rosette Ivy

Potrebbero piacerti anche