You are on page 1of 24

This article was downloaded by: [Western Kentucky University]

On: 23 April 2013, At: 14:32


Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Construction Management and Economics


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/rcme20

Multi-project scheduling: a new categorization for


heuristic scheduling rules in construction scheduling
problems
a
Samir I. G. Allam
a
Business Administration Department, Cairo University, Giza, EgyptProduction and Project
Management, Faculty of Commerce
Version of record first published: 28 Jul 2006.

To cite this article: Samir I. G. Allam (1988): Multi-project scheduling: a new categorization for heuristic scheduling rules in
construction scheduling problems, Construction Management and Economics, 6:2, 93-115

To link to this article: http://dx.doi.org/10.1080/01446198800000010

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should
be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,
proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in
connection with or arising out of the use of this material.
Construction Management and Economics, 1988,6, 93-1 15

Multi-project scheduling: a new categorization


for heuristic scheduling rules in construction
scheduling problems
SAMIR I . G . ALLAM
Production and Project Management, Business Administration Department, Faculty ofcommerce, Cairo
Uniuersity, Giza, Egypt
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

After more than two decades of applying CPM as a planning and scheduling technique there is growing
doubt about some of the advantages initially attributed to it. It is being discarded by both large and small
construction companies, and its use limited to cases where it is required by the clients. Some planners and
schedulers have begun to depend on their own experience.
The author has developed a heuristic planning and scheduling model which uses two new criteria: the
tightness degree of completion due-dates and resource tightness degrees. Throughout the scheduling
experiments, the hypothetical relationship between the performance of the proposed heuristic scheduling
rules and the degrees of completion due-dates and resource tightness proved to be correct and there is a
significant difference between the performance of proposed heuristic scheduling rules in the different
scheduling classes. The new classijication of scheduling problems proved to be practical, effective and
helpful. The scheduling experiments resulted in a new categorizationfor the performance of the proposed
heuristic scheduling rules in the diferent scheduling classes.
Keywords: Planning and scheduling, project management, construction management

Introduction: CPM and the need for a new alternative

Since the developement of critical path methods (CPM), there have been many different
scheduling models based on this technique. Some of these models are mathematical
(analytical) algorithms which aimed to produce optimal scheduling solutions. However,
their application proved to be successful on at most only small projects (Gray, 1981, 103-4;
Kurtulus, 1985; Wiest and Levy, 1977, 106-23); they are therefore good for research and
academic purposes but not for practical and complicated scheduling problems encountered
in construction companies.
Because of this lack of success with the optimization procedures, major efforts have been
expended in developing heuristic scheduling procedures, in which the main objective is to
produce feasible and good solutions. These heuristic scheduling procedures depend on
assigning priorities to scheduling activities using heuristic rules. The performance of these
rules under different conditions, and using different approaches has been tested and reported
in many published papers. Most of the previous studies used the single-project approach in
which separated projects are artificially connected into a large project using dummy
activities. The few other studies (Kurtulus and Davis, 1982), used the multi-project
approach. However, the findings of these studies vary widely.
0144-6193188 $03.00+.12 @ 1988 E. & F.N. Spon Ltd.
There are in existence today hundreds of different heuristic scheduling models based on
CPM and using different heuristic rules. Most of these procedures are kept secret for
proprietary reasons, but none of those which have been discussed in open literature satisfies
all the prerequisites of reliable and applicable scheduling procedures for construction
companies. The main reasons can be summarized as follows:
(1) The scheduling problem in construction companies has been considered by all the
available scheduling models as a normal project scheduling problem. Despite the fact that
the main product of construction companies is one-off, there are some similarities between
this scheduling problem and the job-shop scheduling problem. Because of the dynamic and
continuous nature of work in construction companies, it is normal for new contracts to be
continually awarded, and these contracts need to be considered for scheduling. In fact, there
is no end to the scheduling efforts in construction companies, thus making the scheduling
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

problem more complicated than the normal project scheduling problem.


(2) The available scheduling models have classified the scheduling problems into two main
groups: time-constrained and resource-constrained. Accordingly, the scheduling objective is
significantly different in the two groups. In time-constrained scheduling problems the
objective could be one of the following objectives:
Reducing or minimizing the resource peaks and reducing the fluctuations in demand for
resources through resource levelling.
Minimizing the project durations with minimum increase in total project cost by
expediting the project's most critical activities (time-cost trade-off).
In resource-constrained scheduling problems, the resources required are available only in
limited quantities. Therefore, planners and schedulers have to schedule the activities, if the
required resources are available, and postpone them until the required resources become
available. The scheduling objective is minimizing total project delay. This classification may
be acceptable for normal project scheduling problems or even for construction scheduling
problems from a client point of view. However, in a survey conducted by the author (Allam,
1986,68-79), in which he conducted personal interviews with planners and schedulers in 26
British construction companies, and received another 22 replies by mail, 63% of the replies
indicated that they worked under the pressure of both time and resources and found it was
difficult for them to classify the scheduling problems in either of the two categories. This
finding has been confirmed in another survey conducted by the author, in which he carried
out personal interviews in 15 Egyptian construction companies, and another 11 interviews
with project and site managers in the largest Egyptian construction company (The Arab
Contractors Co.). In the latter survey, 85% of the replies indicated that both time and
resources were scheduling constraints, and they could not classify their scheduling problems
in either of the two categories (Allam, 1986, 266-77).
(3) The other main problem of the available scheduling models is that they are based on
CPM. However, despite the fact that CPM has been available to construction companies for
more than two decades, in a survey published in 1974 (Davis, 1974), 45% of the companies
(the sample contained 400 of the top largest American construction companies), indicated
that they never or seldom used CPM. This was much higher than expected when CPM was
first introduced in the early 1960s. In the author's survey among British construction
companies (Allam, 1986,8&98), 4.9% of the CPM users applied it in all their projects and
68.3% used it in less than half their projects. The fact is that only 34.5% or less of current
Multi-project scheduling 95
CPM users use it in all the phases of project planning and scheduling. Some of the planners
and schedulers whom the author has met - who are chief planners in the largest British
construction companies - have become totally unconvinced by CPM. They have discarded
CPM or limited its application to a few projects where it was absolutely necessary. They have
begun to depend on their own experience and use bar charts as a method of presenting the
developed schedules.

Discarding CPM and using the old-fashioned bar charts instead has been reported before
(Harrison, 1981,78-84). In another informal survey of the American construction industry
(Moder et al., 1983, 174-8), the authors reported that the construction companies varied
widely in the level and style of network use. Small companies did not use the technique at all
or employed scheduling consultants who, according to the same survey, used the technique
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

only when required by the client, which was seldom. In the author's surveys in both Egypt
and the UK, the complexity of CPM is the main application problem. It is the reason for its
rejection and non-acceptance by users, especially site users - CPM is considered a very poor
means of communication. Using CPM together with bar charts, besides the unnecessary
duplication which could lead to errors, especially in companies who do not use computer
facilities (47% of British and 84.6% of Egyptian construction companies), the users would
receive schedules on bar charts which did not show activity interrelationships. This could
lead to wrong decisions, if site managers had to practise their granted authorities and
reschedule some activities for any reasons. Using bar charts together with CPM does not
solve the complexity problem. It may satisfy the need of site users, but CPM has not only
been rejected by site users, but it has also been discarded for the same reason by some
planners and schedulers. The use of sectionized networks (subnetworks), in which planners
and schedulers use a hierarchy of project plans which offer different levels of detail for
different functions and levels of management, is a step in the right direction of simplicity, but
it is not enough to solve the problem and to encourage CPM's potential users. In particular,
the main idea of using networks remains unchanged. Also the idea of using time-scaled
networks or network bar charts (Moder et al., 1983, 170-80) is still complicated and falls
short of providing a comprehensive alternative to CPM for use in construction companies.
The other dimension of the complexity problem of CPM is CPM's arithmetic procedure
and its output. The arithmetic procedure was frequently criticized by planners and
schedulers in both Egyptian and British construction companies throughout the author's
surveys because of its complexity, especially for those companies who do not use computer
facilities in planning and scheduling, and for planners and schedulers who have no academic
qualifications. This problem will be clearer if we consider this arithmetic procedure together
with the associated scheduling algorithms (analytical and heuristic procedures). Despite the
complexity of the arithmetic procedure, the main output of this procedure is activity early
and late start and finish times. This type of output soon becomes out of date if the project
start is delayed for any reasons or the project runs behind schedule. The other type of output
is activity float times. These float times are always subject to misuse, misinterpretation and
confusion because of the large number of float terms (up to seven terms, Lester, 1982,2147).
The author found that very few planners, schedulers, site managers and supervisors, can
understand the differences between them.
The author believes, as do other planners and schedulers in the construction companies,
unless researchers and management consultants find a radical solution to the CPM
application problems, or develop an alternative for it, each manager will have to use his own
scheme, which could involve a limited use of bar charts. This started to happen a few years
ago in some British construction companies who participated in the survey.

A new planning and scheduling model

The author has developed a new planning and scheduling model. This model is based on the
idea - as are all network analysis techniques - of separating planning and scheduling. The
developed model uses one type of diagramming with different levels of detail which can cover
all types of required output, such as plans, work orders and reports. This type of
diagramming takes the familiar shape of bar charts with the advantage of showing activity
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

precedence relationships. The model offers different degrees of depth and scope, which can
satisfy all the different needs in the different project stages; pre-tender planning,
programming, scheduling and resource allocations. The developed model avoids the
excessive use of mathematics, the confusion of early and late start and finish times and the
confusion and misinterpretation of float times. The details of the developed model will be
discussed in another published paper. The distinctive features of the scheduling algorithm in
this model can be summarized as follows:

The model is a heuristic multi-project multi-resource scheduling model. This means that
the model, in assessing the priority of scheduling different activities, depends on heuristic
rules. These rules have been developed to match the general theme of the whole model and
the available data provided by the model in the early stages.
The developed model uses the parallel approach in which planners and schedulers
consider only one time interval, and all activities which could be scheduled (all their
predecessor activities are finished) are ranked according to the used priority rule and
scheduled according to the resource availability; otherwise the activities are postponed to the
next interval, and so forth. The author believes that this approach is more practical for
scheduling multi-project problems in general and for scheduling problems in construction
companies in particular, where there is no end to the scheduling efforts because of the
continuous in and out flow of projects, and the scheduling output should be in work orders
issued to site managers and other departments in the company.
The developed model depends on classifying the type of the scheduling problems using two
new criteria:

The tightness degree of project completion due-dates, which is an indicator to how loose
or tight is the project completion due-date. The author used the allowance factor (a) which
is a contingency value added to project duration to cover different factors of uncertainty
(Allam, 1986,131-6) as an indicator of completion due-date tightness degree. The project
completion due-date was considered very tight if its start was delayed for any reason or the
construction company had to accept the client estimation of project completion due-date
which was shorter than the company's estimation, or the project was already running
behind schedule.
The tightness degree of resource availabilities which reflects how much will be available
of the required resources.
Multi-project scheduling 97
The used heuristic scheduling rules

The author has developed eight new heuristic scheduling rules. They are easy, simple and
calculable rules and based on everyday terms, e.g. completion due-dates, today's date,
activity duration. These rules are calculable and able to be understood and used by all levels,
even supervisory level. A definition of each rule is given in Table 1.

The test hypotheses

There is a difference between the performance of the proposed heuristic scheduling rules
measured by some performance criteria.
There is a relationship between the tightness degree of completion due-dates and the
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

performance of proposed heuristic scheduling rules. This means that one rule performs best
for tighter degrees of completion due-dates and another performs best in looser degrees.
There is a relationship between the performance of proposed heuristic schedulingrules and
the degree of resource tightness.
There is a difference between the performance of proposed heuristic rules in the different
schedulingclasses which reflect different combinations of resource and completion due-dates
tightness degrees.

Project test set

The tightness degree of completion due-dates is defined as the average value of the allowance
factor (a) which is a contingency value added to project duration to cover different factors of
uncertainty (Allam, 1986, 129-36). In these experiments, the value of the allowance factor
ranged between - 0.10 (this means that the completion due-dates are very tight and it may be
that some projects are already running behind schedule) and 0.10 (this means that the
completion due-dates are very loose). This range was divided into five equal intervals with
midpoints -0.10, -0.05,0.00,0.05, 0.10.
The resource tightness degrees are designed to reflect all the possible degrees of resource
availability. They range from 100%, where all the required quantities from all key resources
are available to schedule all activities without delay, to 80% or less in which some of the
required quantities of some or all key resources are not available and where therefore some
activities need to be postponed. This range was divided into five intervals without midpoint
(100%, 95%, 90°h, 85%, 80% or less). Combinations of these two factors give 25 scheduling
classes (cells).
Despite the risk of using artificial projects in these scheduling experiments, the researcher
had to use this type of project because for different reasons, both Egyptian and British
construction companies refused to cooperate and provide samples of their projects. The
project test set contains seven projects with an average of 266 activities. The total number of
activities is 1864. Each activity was allowed to use up to 11 types of resources with no upper
limit for number of resources or the quantities which could be used by any activity. The
project set was replicated 25 times to reflect the different possible combinations of resource
and completion due-dates tightness degrees. This means that the total number of projects
considered in these experiments is 175 with a total number of 46,600 activities and, if we
98

Table 1. The proposed heuristic scheduling rules.


1 The criticality indicator (CRZ):

where
d(ijm)= the completion due date for activity (m)
E(ijm)= the duration of activity (m)
D = today's date
i =the project notation
j = the main activity notation
m= the activity notation
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

2 The earliest completion due-date (EDD):


EDD(ijm)= min[d(ijm)]
3 The earliest modified completion due-date:
+
MDD(ijm)=d(ijm)+ max{O, [ ( D E(ijm))-d(ijm)]}
4 The minimum slack (MZNSLK):
MINSLK(ijm)= [(d(ijm)- D ) - E(ijm)]
5 The minimum proportional slack:

6 The total work content (TWV):

where
C ( l ) =the unit cost of resource ( 1 )
R(l)= the daily quantity used from resource ( I ) for the activity.
7 The weighted criticality indicator ( WCRZ):
There are two ways used to weigh the criticality indicator;

where
N = the number of projects in the scheduling set
R =Value given to every project in the scheduling set according to its importance to the
company; R = N if the project is very important, and R= 1 when the project is the least
important project in the set. The value of R depends on indicators, some of which are
quantified, e.g. the rate of delay penalty, or qualitified indicators, e.g. the management
judgement, the client importance, and so on.
r
WCRZ2(ijm)= CRI(ijmim) 1 -

where
ACF(ijm)= the amount of cash inflow if activity m completed
Multi-project scheduling 99
consider that there are eight heuristic rules tested and up to eleven types of resources
considered for every activity, then the number of activities is 4,100,000.
The types of activities allowed in these experiments are pre-emptive and non-pre-emptive
activities. The non-pre-emptive activities once started should be completed, but pre-emptive
activities could be interrupted and rescheduled later when the required resources become
available.

The statistical tests used

The researcher used two non-parametric tests for the following reasons. First, there is no
evidence that the project delays are normally distributed in the real-world scheduling
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

problems (Kurtulus and Davis, 1982). Secondly, the project test set has been replicated 25
times to reflect the different degrees of completion due-dates and resource tightness. This
means that the project samples are not independent, and independence is one of the main
assumptions of the parametric tests. Finally, the researcher used two of the most powerful
non-parametric tests (Siegle, 1956), which are based on more general assumptions and hence
would also be true under the more restrictive assumption of normality. The tests used are:
Friedman two-way analysis of variance, to test the significance of the difference between
the performance of proposed heuristic scheduling rules and the effect of the tightness degrees
on the performance of the proposed rules.
Wilcoxon signed-rank test, to categorize the performance of the proposed rules in the
different scheduling classes. In this test, the performance of the first best rule was tested
against the performance of the second best rule. If the difference was significant, the first rule
was judged as the best rule in this scheduling cell. The performance of the second best rule
was then tested against the performance of the third best rule, and the test continued.
However, if the difference between the performance of the two rules was insignificant, the two
rules were tied together and either of them could be used and the test continued until all rules
in all scheduling classes were tested and categorized.

Performance criteria used

The following criterion functions are used in assessing the efficacy of the heuristic scheduling
rules examined in these scheduling experimentations. The choice of an appropriate objective
function may differ in various scheduling environments and in different periods of time.
Total project delay is the difference between project completion due-date assigned in
project contract and actual completion date. Project actual completion date is the
completion date for the last scheduled activity.
Resource utilization rate is the percentage of the quantity used against the available
quantity of one type of resource. The researcher calculated the daily utilization rates for all
types of key resources used in these experiments. To compare between the performances of
the proposed rules, the researcher used the average utilization rate.
Average cost of project delay may include one or more of the following items:
Direct cost of dealing with customer, paper work, telephone calls, and so on.
Penalty clause assigned in project contract which may be on a daily or weekly basis.
Loss of goodwill, which may mean losing the customer for some or all future contracts, and
a damaged reputation which could turn other customers away, especially in highly
competitive markets.
Expediting cost.
Normally the aggregate of these items of cost is monotonic increasing with project delay. In
this research, the author calculated the cost of project delay as total project delay multiplied
by the penalty rate assigned in project contract.

Main experimentation results

( 1 ) The diflerence between the performance of proposed heuristic scheduling rules


Downloaded by [Western Kentucky University] at 14:32 23 April 2013

The average project delay varies between 18.8 days and 41.4 days depending upon the
heuristic rule used. The average utilization rate varies between 0.360 and 0.377. Also, the
average cost of project delay varies between £2671 and £6292 depending upon the heuristic
rule used. Applying Friedman two-way analysis of variance, the difference between the
performance of proposed rules is significant at level 0.001. Therefore, the first hypothesis
about the difference between the performance of proposed rules proved to be correct. There is
a significant difference between the performance of these rules, measured by any of the
criteria used.
At this stage, we can categorize the performance of the proposed rules using the Wilcoxon
signed-rank test. Table 2 presents three categorizations according to the three performance
criteria used in the experiments. MZNSLK is the best rule by any performance criteria. The
difference between its performance and the performance of the other rules is significant at
levels between 0.025 or over. It is worth noting at this stage that some rules, which had been
reported in previous studies as the best rules, performed poorly in these experiments, e.g.
MDD (MOD).This rule had been reported in much job-shop schedulingliterature as the best
heuristic rule (e.g. Baker and Kenet, 1983,1984;Baker, 1984).In these experiments it came in
sixth place according to two performance criteria used and in fifth place according to
resource utilization rate.

Table 2. The categorization of the performance of the proposed heuristic scheduling


rules.
The average cost
of project delay The total project delay The average utilization rate
1-MINSLK" 1-MINSLK" 1-MINSLK"
2-PMNSLK 2-PMNSLK 2-PMNSLK
WCRI2 WCRI2 WCRI2
WCRIl WCRIl WCRIl
3-CRIb 3-CRZ CRI
4-EDD" 4-EDD" 3-MDD"
MDD MDD 4-TWV
TWV TWV SEDD
- -- -- - -

The difference between the performance of the rules is significant at level 0.05.
"significant at level 0.025 or over.
bsignificant at level between 0.10 and less than 0.05.
Multi-project scheduling 101
( 2 ) The relation between the performance of proposed heuristic rules and the completion
due-dates tightness degrees
First, Fig. 1 displays the relationship between the tightness degrees of completion due-dates
and the performance of proposed heuristic scheduling rules measured by average project
delay. The effect of the tightness degrees of completion due-dates is significant and clear: the
tighter the completion due-dates, the longer the average project delay. MINSLK
outperformed the other rules in all the tightness degrees. The performance of MDD, EDD
and TWV, which had been reported as good rules in previous studies, deteriorated as the
completion due-dates became tighter, and came in the last three places. The difference
between WCRII, WCRI2, CRI and PMNSLK is insignificant, especially in the tighter
degrees of completion due-dates, and they form one set and any of them could be used.
Secondly, Fig. 2 displays the performance of proposed heuristic scheduling rules,
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

measured by the average utilization rate, in the different tightness degrees of completion due-
dates. The effect of completion due-dates tightness degrees on the performance of proposed
rules is significant at levels between 0.10 and 0.05. Despite that the difference between the
performance of the proposed rules, measured by average utilization rate in the different
degrees of completion due-dates tightness, proved to be insignificant. In Fig. 2 we can note
that MDD and EDD performed better than other rules in looser degrees of completion due-
dates tightness and their performance deteriorated as the completion due-dates became
tighter. MZNSLK performed less efficiently than MDD and EDD in looser degrees and
outperformed the other rules in the tighter degrees.
Finally, Fig. 3 shows graphically the difference between the performance of proposed
heuristic rules, measured by average cost of project delay in the different degrees of
completion due-dates. The main finding is quite similar to previous findings about the effect
of the tightness degrees of completion due-dates on the performance of proposed rules and
the significant difference between the performance of the proposed rules in the different
degrees of completion due-dates tightness. The other finding is the poor performance of
MDD, EDD and TWV and the encouraging performance of MINSLK in all tightness
degrees.

( 3 ) The relationship between the performance of proposed heuristic rules and the degrees of
resource tightness
First, Fig. 4 shows the performance of proposed heuristic rules, measured by average project
delay in the different degrees of resource tightness. Fig. 5 presents the same relationship with
one difference: in this figure we excluded the tightest degree of resource availability (80% or
less). The effect of the resource tightness degrees on the performance of proposed rules is
significant and clear in the two figures. The tighter the resource, the longer the average of
project delay. MINSLKoutperformed all the other rules in all degrees of resource tightness.
TWV, which had been reported as a good rule (Kurtulus and Davis, 1982),is the poorest rule
in the first four degrees of resource tightness. MDD and EDD performed poorly in all degrees
of resource tightness.
Secondly, these last two findings on the effect of resource tightness and the significant
difference between the performance of proposed rules, have been confirmed by studying the
performance of the proposed rules, measured by both average utilization rate and average
cost of project delay, as shown in Figs 6 and 7.
Downloaded by [Western Kentucky University] at 14:32 23 April 2013
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

Multi-project scheduling
Downloaded by [Western Kentucky University] at 14:32 23 April 2013
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

Legend
A CRI

-----
WCRll
WCRIZ
8 MDP_-------
X TWV
--
O MINSM
-----
CB PMNSLK

Fig. 4. The effect of the resource tightness degrees on heuristic scheduling rule performance.
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

Legend
I A CRI
---
X EDD
-----
WCRll
[El WCRB . -
XX MDD
-----------*

x TWV

@ ----
PMNSLK

Resource Tightness Degrees


Fig. 5. The effect of the resource tightness degrees on heuristic scheduling rule performance, excluding the tightest degree.
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

The Average of The Res.Utilisation Rates


Downloaded by [Western Kentucky University] at 14:32 23 April 2013
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

Multi-project scheduling
( 4 ) Categorization of the performance of proposed heuristic rules in diferent combinations
of completion due-dates and resource tightness
So far, we have statistically proved the significance of the effect of both completion due-dates
and resource tightness degrees on the performance of proposed heuristic scheduling rules,
measured by three different performance criteria: average project delay, average resource
utilization rate, and average cost of project delay. We have also proved that the performance
of the proposed heuristic rules varied significantly throughout the different degrees of
scheduling tightness.
However, it is understood that any scheduling case contains a combination of both
completion due-dates and resource tightness degrees. In the experiments conducted, the
tightness of completion due-dates and resource availability had been divided into five
degrees each. This means that there were 25 different scheduling cases. Fig. 8 presents as an
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

example, the performance of the proposed rules measured by average project delay in the
different scheduling cases which reflect 25 possible combinations of both completion due-
dates and resource tightness degrees. The effect of the tightness degrees on the performance of
the rules is significant and clear. However, we can note that the performance of the proposed
rules has been affected by resource tightness more than by completion due-dates tightness.
We can come to the same conclusion if we study the performance of these rules measured by
average utilization rate and average cost of project delay.
Applying the Friedmann two-way analysis of variance to the data available on the
performance of proposed heuristic scheduling rules in the 25 scheduling cases, measured by
any of the performance criteria used, provides further evidence to support the previous
conclusions:
The performance of proposed heuristic rules is significantly affected by the scheduling
tightness degrees.
The difference between the performance of proposed heuristic rules, measured by any of
the performance criteria used, proved to be significant at levels between 0.005 and 0.001.
The latter two conclusions encouraged the researcher to categorize the performance of the
proposed heuristic rules into the 25 different scheduling cases. The Wilcoxon signed-rank test
was used, as explained before, to test the significance of the difference between the proposed
rules. The rules were put in one set if the difference between their performance was not
significant at an acceptable level (at least 0.10). The use of three performance criteria required
three categorizations. Tables 3-5 present the categorizations of the performance of the
proposed heuristic rules in the different scheduling cases.

Conclusions

The performance of proposed heuristic scheduling rules is significantly affected by both


completion due-dates and resource tightness degrees.
The performance of proposed heuristic scheduling rules varied significantly throughout
the scheduling cases, which contained different degrees of completion due-dates and resource
tightness. The difference between the performance of these rules is significant at levels
between 0.025 and 0.001, no matter what performance criterion was used.
There are some rules, which were reported in previous studies in both job-shop scheduling
Table 3. The categorization of the performances of the proposed heuristic scheduling rules (the total project delay)*.
Resource The completion due-date tightness degrees
tightness
degrees
I-CRI, EDD I-CRI, EDD l-CRI, WCRIl I-CRI, WCRIl 1-MINSLK
WCRI 1 MDD, WCRIl WCRI2 PMNSLK PMNSLK
WCRI2 WCRI2 MINSLK MINSLK WCRIl
MDD, MINSLK MINSLK PMNSLK 2-WCRI2 CRI
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

PMNSLK PMNSLK 2-MDD 3-MDD 2-WCRI2


2-TWV 2-TWV 3-EDD 4EDD 3-MDDb
l-CRI, EDD I-CRI, EDD 1-CRI 1-MINSLK 1-MINSLK
WCRI1 WCRI1 PMNSLK PMNSLK CRI
WCRI2 WCRI2 MINSLK 2-WCRI2 2-WCRI 1
MDD, MINSLK MDD, MINSLK 2-WCRI1 3-CRIb PMNSLK
PMNSLK PMNSLK WCRI2 4-WCRIb WCRI2
2-TWV 2-TWV 3-MDD", EDD 5-MDD 3-MDD
4-TWV
I-CRI, EDD 1-CRI 1-MINSLK 1-MINSLKb 1-MINSLK
WCRI1 MINSLK 2-WCRI, CRI 2-CRI 2 4 R 1 , WCRI1
WCRI2, MDD WCRI1 3-WCRI 1 PMNSLK WCRI2
MINSLK WCRI2 PMNSLK 3-WCR12b PMNSLK
2-TWV PMNSLK 4EDD WCRI1 3-MDD
2-EDD, MDD 4-MDD, TWV
1-EDD, WCRI2 1-MINSLK 1-MINSLK 1-MINSLK 1-MINSLK
MDD 2-WCRI2 PMNSLK 2-WCRI 1 2-PMNSLK
MINSLK 3-WCRI1, CRI 2-WCRI2, CRI 3-WCRI2 CRI, WCRIl
2-CR1, WCRIl PMNSLK 3-WCRI 1, EDD PMNSLK 3-WCRI2 .
PMNSLK 4-EDD 4-MDDb 4-CRI" MDD
3-TWV
1-MINSLK 1-MINSLK 1-MINSLKb 1-MINSLK I-MINSLK
2-CRI 2-WCRI1 2-WCRI1 2-PMNSLK 2-PMNSLK
80% 3-WCRI2 WCRI2 PMNSLK 3-WCRI2 TWV, CRI
or less PMNSLK 3-PMNSLK 34RI 4-CRIb WCRI1
4-WCRI1 4-CRIb 4-WCR12b WCRI1 3-WCRI2
EDD EDD TWV TWV
*Significant at level 0.05.
"Significant at level between 0.05 and 0.001.
bSignificant at level between 0.10 and 0.05.
Table 4. The categorization of the performances of the proposed heuristic scheduling rules (the average utilization rate)*.
Resource The completion due-date tightness degrees
tightness
degrees 5% 0% (5%) (10%)
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

l-CRI, EDD I-CRI, EDD l-CRI, WCRIl l-CRI, WCRI1 1-MINSLK


WCRI 1 MDD, WCRIl WCRI2 PMNSLK PMNSLK
WCRI2 WCRI2 MINSLK MINSLK WCRI 1
MDD, MINSLK MINSLK PMNSLK 2-WCRI2 CRI
PMNSLK PMNSLK 2-MDD 3-MDD 2-WCRI2
2-TWV 2-T WV 3-EDD 4-EDD 3-MDDb
l-CRI, EDD l-CRI, EDD l-CRI, WCRI2 1-MINSLK 1-MINSLK
MDD, MINSLK MDD, MINSLK MINSLK PMNSLK CRI
PMNSLK WCRI1 PMNSLK 2-WCRI2, CRI 2-WCRI1
WCRI 1 WCRI2 2-WCRI1 3-WCRI1 PMNSLK
WCRI2 2-TWV 3-MDD, EDD 4-MDD 3-WCRI2
2-TWV 4-MDDb
1-EDD, MDD l-CRI, WCRIl 1-MINSLK 1-MINSLK 1-MINSLK
CRI, WCRIl WCRI2 2-WCRI2, CRI 2-CRIb 2-CR1, WCRIl
WCRI2 MINSLK 3-WCRI1 PMNSLK PMNSLK
MINSLK 2-EDD, MDD PMNSLK 3-WCR12b WCRI2
PMNSLK 3-TWV 4-EDD WCRIl MDD
1-EDD, MDD 1-MINSLK 1-MINSLK 1-WCRI1 1-MINSLK
WCRI2 2-WCRI2 PMNSLK MINSLK 2-CRI
MINSLK 3-WCRI1 2-WCRI2, CRI 2-WCRI2 3-PMNSLKb
2-CR1, WCRIl 4-CRIb 3-WCRI1 PMNSLK WCRI1
PMNSLK PMNSLK EDD 3-CRI 4-WCRI2, MDD
1-MINSLK 1-MINSLK~ 1-MINSLK 1-MINSLK
80% 2-CRI 2-WCRI 1 2-PMNSLK 2-PMNSLK
or less 3-WCR12b PMNSLK WCRI2 3-WCRIl
PMNSLK 3-CR1, WCRI2 3-CRIb CRI, TWV
4-EDD, WCRI2
*Significant at level 0.05.
4-CRI, EDD TWV WCRI1, TWV WCRI2
-
?k
G
t
bSignificant at level between 0.10 and 0.05.
Table 5. The categorization of the performances of the proposed heuristic scheduling rules (the average cost of project delay)*.
Resource The completion due-date tightness degrees
tightness
degrees 10% 5% 0% (5%) (10%)
ILCRI, EDD ILCRI, EDD I-CRI, WCRIl 1LCR1, WCRI2 1-MINSLK
WCRI1 MDD, WCRIl WCRI2 PMNSLK PMNSLK
WCRI2 WCRI2 MINSLK MINSLK WCRI1
100% MDD, MINSLK MINSLK PMNSLK 2-WCRI 1 2-CRI
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

PMNSLK PMNSLK 2-MDD 3-MDD 3-WCRI2


2-TWV 2-TWV 3-EDD 4-MDDb
l-CRI, EDD ILCRI, EDD l-CRI, WCRI2 1-MINSLK 1-MINSLK
WCRI 1 WCRI 1 PMNSLK PMNSLK 2-CR1, WCRIl
WCRI2 WCRI2 MINSLK 2-WCRI2, CRI
95% MDD, MINSLK MDD, MINSLK 2-WCRI1 3-WCRIlb PMNSLK
PMNSLK 2-TWV 3-MDD" 4-MDD WCRI2
2-TWV MDD
1-MDD, EDD 1-CRI 1-MINSLK 1-MINSLK 1-MINSLK
WCRI1 WCRIl 2-WCR12, CRI 24RI 2-CR1, WCRIl
WCRI2, CRI WCRU 3-WCRI 1 PMNSLK WCRI2
90% MINSLK MINSLK PMNSLK 3-WCR12b PMNSLK
2-PMNSLK 2-EDD, M D D 4-EDD WCRI 1 3-MDD
3-TWV
1-EDD, WCRI2 1-MINSLK 1-MINSLK 1-MINSLK 1-MINSLK
MDD 2-WCRI2 PMNSLK WCRI1 2LCRI
85% MINSLK 3-WCRI 1 2-WCRI2, CRI 2-WCRI2 3-PMNSLKb
2-CR1, WCRIl 4CRIb 3-WCRI1 PMNSLK 3-WCRI 1
PMNSLK PMNSLK EDD 4CRI 4-WCRI2, MDD
1-MINSLK 1-MINSLK 1-MINSLK~ 1-MINSLK 1-MINSLK
2-CRI 2-WCRI1 2-WCRI1 2-PMNSLK 2-PMNSLK
80% WCR12b WCRI2 PMNSLK WCRI2 3-WCRIlb
or less PMNSLK 3-PMNSLK 3-CRI CRI~ CRI, TWV
4-WCRI2 4CRI WCRI2 WCRI1 WCRI2
EDD EDD TWV TWV
*Significant at level 0.05.
bSignificant at level between 0.10 and 0.05.
and project scheduling literature as the best heuristic scheduling rules, and which performed
poorly throughout these scheduling experimentations: e.g. MDD, EDD and TWV.
MINSLK in general outperformed all the other proposed heuristic scheduling rules but
this does not mean that we should recommend its application in all scheduling cases, as
previous researchers did. We can simply add the following notes about its application:
When resource availabilities are very tight MINSLK is the best heuristic rule, no matter
what performance criteria are used.
MINSLKis the best heuristic scheduling rule when completion due-dates are very tight. It
comes first in all the three categorizations with only one exception. Even in this case the
difference between the performance of CRI, WCRII and MINSLK is insignificant.
In cases when resource availabilities are loose, and completion due-dates are less tight, the
choice between proposed heuristic scheduling rules should depend on the scheduling
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

objectives.
The proposed model proved throughout these scheduling experiments to be applicable,
practical, flexible and efficient.
The proposed model is a dynamic scheduling model. It accepts any number of projects
with any number of activities, and any number of projects can be added to the scheduling set
as soon as contracts are received. There is no upper limit to the number of resources. The
heuristic scheduling rules can be changed at any time, when any of the control parameters
(completion due-dates and resource tightness degrees) are changed, or when the scheduling
objective is changed.

References

Allam, S.I.G. (1986) Planning and Scheduling: A New Model for Planning and Scheduling
Construction Projects, Unpublished Ph.D., Henley: The Management College, and Brunel
University.
Baker, K.R. (1984) Sequencing rules and due-dates assignments in a job-shop, Management Science,
30,9,1093-1104.
Baker, K.R. and Bertrand, J.W.M. (1981) An investigation of due-date assignment rules with
constrained tightness, Journal of Operations Management, 1, 3, 109-20.
Baker, K.R. and Bertrand, J.W.M. (1982) A dynamic priority rule for scheduling against due-dates,
Journal of Operations Management, 3, 1 , 3 7 4 2 .
Baker, K.R. and Kanet, J.J. (1983) Job shop scheduling with modified due-dates, Journal of Operations
Management, 4, 1, 11-22.
Baker, K.R. and Kanet, J.J. (1984) Improved decision rules in a combined system for minimizing job
tardiness, International Journal of Production Research, 22, 6, 917-21.
Davis, E.W. (1974) CPM use in top 400 construction firms, Proceedings ofthe American Society of Civil
Engineers, 100, COI, March 1981, 39-49.
Davis, E.W. and Patterson, J.H. (1975) A comparison of heuristic and optimum solutions in resource-
constrained project scheduling, Management Science, 21, 8, 944-55.
Gray, C.F. (1981) Essentials of Project Management, Petrocelli Books, UK.
Harrison, F.L. (1981) Advanced Project Management. Gower, UK.
Kurtulus, I. (1985) Multi-project scheduling: analysis of scheduling strategies under unequal delay
penalties, Journal of Operations Management, 5, 3, 291-307.
Kurtulus, I. and Davis, E. W. (1982) Multi-project scheduling: categorization of heuristic scheduling
rules performance, Management Science, 28, 2, 161-72.
Multi-project scheduling 115
Lester, A. (1982) Project Planning and Control, Butterworths, London.
Meredith, J.R. and Mantel, S.J. (1985) Project Management, A Managerial Approach. Wiley, New
York.
Moder, J.J., Phillips, C.R. and Davis, E. (1983) Project Management with C P M , PERTand Precedence
Programming, 3rd edn. Van Nostrand Reinhold, New York.
Siegle, S. (1956) Non-Parametric Statistics for the Behavioral Sciences. McGraw-Hill, pp. 75-83,
166-172.
Wiest, J.D. and Levy, F.K. (1977)A Management Guide to PERTICPM, with GERTIPDMIDCPMand
Other Networks, 2nd edn. Prentice-Hall, Englewood Cliffs, NJ.
Downloaded by [Western Kentucky University] at 14:32 23 April 2013

You might also like