See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228904075

Optimal Planning and Scheduling for Repetitive Construction Projects

Article  in  Journal of Management in Engineering · January 2006

DOI: 10.1061/(ASCE)0742-597X(2006)22:1(11)

CITATIONS READS
141 3,518

3 authors, including:

Khaled Hesham Hyari Khaled El-Rayes

Hashemite University University of Illinois, Urbana-Champaign
55 PUBLICATIONS   701 CITATIONS    155 PUBLICATIONS   3,720 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Research Report No. FHWA-ICT-14-001 View project

project schedule optimization models View project

All content following this page was uploaded by Khaled El-Rayes on 04 June 2014.

The user has requested enhancement of the downloaded file.

 Optimal Planning and Scheduling for Repetitive
Construction Projects
Khalied Hyari1 and Khaled El-Rayes, M.ASCE2

Abstract: This paper presents a multiobjective optimization model for the planning and scheduling of repetitive construction projects.
The model enables construction planners to generate and evaluate optimal construction plans that minimize project duration and maximize
crew work continuity, simultaneously. The computations in the present model are organized in three major modules: scheduling, optimi-
zation, and ranking modules. First, the scheduling module uses a resource-driven scheduling algorithm to develop practical schedules for
repetitive construction projects. Second, the optimization module utilizes multiobjective genetic algorithms to search for and identify
feasible construction plans that establish optimal tradeoffs between project duration and crew work continuity. Third, the ranking module
uses multiattribute utility theory to rank the generated plans in order to facilitate the selection and execution of the best overall plan for
the project being considered. An application example is analyzed to illustrate the use of the model demonstrate its new capabilities in
optimizing the planning and scheduling of repetitive construction projects.
DOI: 10.1061/共ASCE兲0742-597X共2006兲22:1共11兲
CE Database subject headings: Road construction; Highway construction; Scheduling; Resource management; Optimization;
Evolutionary computation; Computer models; Construction management.

Introduction Available planning and scheduling models that focused on

minimizing the duration of repetitive construction projects can be
Examples of repetitive construction projects include highways, grouped into two main categories: 共1兲 models that provide strict
tunnels, bridges, railways, pipeline networks, sewer mains, high-
compliance with crew work continuity 共Selinger 1980兲; and 共2兲
rise buildings, and housing development projects. In this class of
models that allow interruptions to crew work continuity 共Russell
projects, construction crews are often required to repeat the same
work in various locations of the project, moving from one loca- and Caselton 1988; El-Rayes 2001兲. All these models are capable
tion to another. Due to this frequent crew movement, available of generating a single optimal solution that minimizes the dura-
scheduling methods for repetitive construction projects focus on tion of the project being considered. For example, Fig. 1
maximizing crew work continuity by enabling each crew to finish shows the single optimal solution produced by one model from
work in one location of the project and move promptly to the next each of the above two categories. Although both models analyzed
in order to minimize work interruptions. The application of work the same application example, the first category model 共Selinger
continuity improves the overall productivity of construction crews 1980兲 that did not allow interruptions produced a project duration
due to: 共1兲 minimizing their idle time during their frequent move- of 117.9, while the second category model 共El-Rayes and Moselhi
ments on site; and 共2兲 maximizing their benefits from learning 2001兲 that allowed 15 days of interruptions generated a project
curve effects 共Ashley 1980; El-Rayes 2001兲. Despite the advan- duration of 106.8 as shown in solutions A and B in Fig. 1,
tages of crew work continuity, its strict application can lead to a respectively.
longer overall project duration 共Selinger 1980; Russell and Casel- The two schedules shown in Fig. 1 illustrate two possible
ton 1988; El-Rayes and Moselhi 2001兲. This led to a number of tradeoffs between the two important and conflicting objectives of
research studies that investigated the impact of 共Russell and Ca- minimizing the project duration and maximizing crew work con-
selton 1988; Senouci and Eldin 1996; crew work continuity on
tinuity. As stated earlier, maximizing crew work continuity can
the planning and scheduling of repetitive construction projects;
produce significant improvements in construction productivity
Harris and Ioannou 1998; El-Rayes 2001; El-Rayes and Moselhi
2001; Hegazi and Wassef兲. and reductions in crew utilization costs, however it can also lead
to a longer overall project duration and an increase in project
1 overhead costs. As such, construction planners need to evaluate
Assistant Professor, Dept. of Civil Engineering, The Hashemite
Univ., P.O. Box 150459, Zarqa 13115, Jordan. E-mail: [email protected]
all these feasible tradeoffs in order to select a construction plan
2
Assistant Professor, Dept. of Civil and Environmental Engineering, that strikes an optimal balance between minimizing the project
Univ. of Illinois at Urbana-Champaign, Urbana, IL 61801 共corresponding duration and maximizing its crew work continuity. Available
author兲. E-mail: [email protected] planning and scheduling models for repetitive construction are
Note. Discussion open until June 1, 2006. Separate discussions must incapable of generating these feasible tradeoffs as they can only
be submitted for individual papers. To extend the closing date by one produce a single optimal solution as shown in Fig. 2. There is a
month, a written request must be filed with the ASCE Managing Editor.
pressing need for advanced models that can help construction
The manuscript for this paper was submitted for review and possible
publication on February 2, 2005; approved on May 9, 2005. This paper is planners in generating and evaluating all the feasible tradeoff so-
part of the Journal of Management in Engineering, Vol. 22, No. 1, lutions shown in Fig. 2 in order to select an optimal plan that
January 1, 2006. ©ASCE, ISSN 0742-597X/2006/1-11–19/$25.00. satisfies the specific requirements of the project being considered.

JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / JANUARY 2006 / 11

 Fig. 1. Minimum duration solution provided by existing models

Objective module that develops practical schedules for repetitive construc-

tion projects; 共2兲 a multiobjective optimization module that
The objective of this paper is to present the development of a searches for and identifies near optimal tradeoffs between project
multiobjective optimization model for planning and scheduling of duration and crew work continuity; and 共3兲 a ranking and selec-
repetitive construction projects. The model provides the capability tion module that ranks the generated solutions according to the
of: 共1兲 generating a set of feasible construction plans that estab- specific requirements of the project being considered 共see Fig. 3兲.
lish optimal tradeoffs between project duration 共D*兲 and crew The following sections present a detailed description of these
work continuity 共R*兲, as shown in Fig. 2; 共2兲 identifying an opti- three major modules.
mal level of resource utilization for each construction activity 共i兲
in the project; and 共3兲 ranking the generated optimal plans to
facilitate the selection of the best overall plan for executing the Scheduling Module
project 共see Fig. 3兲. In order to provide this output, the present
model requires construction planners to input readily available The main objective of this module is to develop practical sched-
data including: 共1兲 project data that specifies the number of re- ules for repetitive construction projects. The module is designed
petitive activities in the project 共I兲 and the job logic and interre- to provide 共1兲 flexibility in considering typical and nontypical
lationships among them; and 共2兲 activity data that specifies the repetitive activities that have identical and varying durations in
number of repetitive sections 共J兲, the quantity of work 共Qi,j兲 of different repetitive units 共Selinger 1980; Russell and Caselton
activity 共i兲 in each unit 共j兲, available crew formations 1988; Moselhi and El-Rayes 1993兲; and 共2兲 practicality in com-
共ni = 1 – Ni兲 that can perform each activity 共i兲 and their daily plying with all scheduling constraints, including crew work con-
productivity rates 共Pi,n兲, as shown in Fig. 3. tinuity, crew availability, and job logic constraints 共El-Rayes and
The planning and scheduling computations in the present Moselhi 1998; El-Rayes 2001兲. The module is also designed to
model are organized in three major modules: 共1兲 a scheduling compute: 共1兲 the scheduled start 共Si,j兲 and finish 共Fi,j兲 dates of

Fig. 2. Tradeoffs between project duration and crew work continuity

12 / JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / JANUARY 2006

 Fig. 3. Optimization model for scheduling repetitive construction

construction for each activity 共i兲 in each repetitive unit 共j兲; 共2兲 the 2. Calculate the earliest start 共SCrew关i,j兴兲 and finish 共FCrew关i,j兴兲
total project duration 共D兲; and 共3兲 the total number of crew inter- times for activity 共i兲 in unit 共j兲 that satisfy crew availability
ruption days 共R兲 at the project level. To this end, the scheduling and crew work continuity constraints, assuming that the ac-
computations in this module are performed using the following tivity has no predecessors and accordingly its first repetitive
eight major steps 共see Fig. 4兲: unit 共j = 1兲 can start at time zero as shown in Eq. 共2兲. SCrew关i,j兴
1. Calculate the construction duration 共di,j兲 of activity 共i兲 in
and FCrew关i,j兴 are calculated using Eqs. 共3兲 and 共4兲 that con-
each repetitive unit 共j兲, using its quantity of work 共Qi,j兲 and
the daily productivity rate 共Pi,n兲 of the selected crew option sider 共1兲 the availability the crew after the completion of
as shown in Eq. 共1兲 and the example foundation activity in work in its previously assigned repetitive unit 共FCrew关i,j−1兴兲;
Fig. 4 and 共2兲 the specified crew interruption time applied after the
completion of the previous unit j − 1 共Interi,j−1兲, as shown in
Qi,j Fig. 4. Although the computations of SCrew关i,j兴 and FCrew关i,j兴
di,j = 共1兲
Pi,n are formulated to satisfy crew availability and crew work

Fig. 4. Scheduling computations for foundation activity

JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / JANUARY 2006 / 13

 continuity constraints, they do not comply with the job logic plies with crew availability and work continuity 共SCrew关i,j兴兲 as
constraint. This constraint requires that activity i 共e.g., foun- shown in Eq. 共5兲. This time difference represents the required
dation兲 can start only after the completion of its predecessor delay in SCrew关i,j兴 and FCrew关i,j兴 to further satisfy the job logic
activity i − 1 共e.g., excavation兲 in each repetitive unit 共j兲. As constraint, as shown in Fig. 4
such, the following three steps are designed to shift SCrew关i,j兴
and FCrew关i,j兴 to a later date in order to further comply with ⌬i,j = SLogic关i,j兴 − SCrew关i,j兴 共5兲
the job logic constraint 共see Fig. 4兲: 5. Identify the maximum time difference from those calculated
SCrew关i,1兴 = 0 共2兲 in the previous step 共Shifti兲 using Eq. 共6兲. As shown in Fig. 4,
this time shift is then used to delay SCrew关i,j兴 and FCrew关i,j兴 in
order to identify the scheduled start 共Si,j兲 and finish 共Fi,j兲
SCrew关i,j兴 = FCrew关i,j−1兴 + Interi,j−1 共3兲
times that satisfy all scheduling constraints for activity 共i兲 in
each repetitive unit 共j兲 using Eqs. 共7兲 and 共8兲
FCrew关i,j兴 = SCrew关i,j兴 + di,j 共4兲
Shifti = MaxJj=1关⌬i,j兴 共6兲
3. Identify the earliest start time 共SLogic关i,j兴兲 that satisfies the job
logic and precedence relationships between activity 共i兲 and
Si,j = SCrew关i,j兴 + Shifti 共7兲
all its predecessors. These relationships can be either finish
to start, start to start, finish to finish or start to finish with or
without lag time. For example, if the precedence relationship Fi,j = FCrew关i,j兴 + Shifti 共8兲
between two activities is finish to start, then the successor
6. Identify the total crew interruption days in activity i 共Interi兲
activity 共i兲 in each unit 共j兲 can start only after the completion
using the following equation:
of its predecessor activity 共i − 1兲 and the lag time 共lagi,i−1兲
between the two activities 共i.e., SLogic关i,j兴 艌 Fi−1,j + lagi,i−1兲. J
4. Calculate the time difference 共⌬i,j兲 between the earliest start
time that satisfies job logic 共SLogic关i,j兴兲 and that which com-
Interi = 兺
j=1
Interi,j 共9兲

Fig. 5. Multiobjective optimization module

14 / JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / JANUARY 2006

7. Compute the total project duration 共D兲 by identifying the In order to search for nondominated construction plans that
time difference between the scheduled finish time of the last specify the optimal values for the aforementioned two resource
activity in its last repetitive unit 共FI,J兲 and the start of the first utilization variables, the present multiobjective optimization
activity in the first unit 共S1,1兲, as shown in the following module is implemented using a multiobjective genetic algorithm
equation: 共Zitzler and Thiele 1999; Deb et al. 2000; Coello and Lechuga
2001; El-Rayes and Kandil 2005; El-Rayes and Khalafallah 2005;
D = FI,J − S1,1 共10兲 Kandil and El-Rayes 2005兲. This algorithm adopts the concept of
8. Calculate the total number of interruption days 共R兲 for all Pareto optimality to support multiobjective optimization, and the
crews in the project using the following equation: survival of the fittest criteria to evolve solutions over generations
in order to yield near optimal solutions. The present module starts
I the optimization process by randomly generating a number of
R= 兺 Interi 共11兲 initial resource utilization plans 共k = 1 – K兲, where each represents
i=1 a random selection of a crew formation option 共ni兲 and a crew
interruption vector 共vi兲 for each repetitive activity 共i兲 in the
project. These randomly generated plans form the parent popula-
tion 共Pt兲 of the first generation 共t = 1兲 that evolves into a near
Multiobjective Optimization Module
optimal solution after a number of predetermined generations 共T兲.
The primary purpose of this module is to search for and identify As shown in Fig. 4, this optimization process is performed using
a set of optimal construction plans that simultaneously minimize the following four major cyclical steps:
project duration and maximize crew work continuity for repetitive 1. Calculate the project duration 共Dk兲 and total interruption
construction projects. Each of these identified optimal plans pro- days 共Rk兲 for each resource utilization plan 共k = 1 – K兲 in the
vides a unique and nondominated tradeoff between project dura- first generation 共t = 1兲, using the earlier described scheduling
tion and crew work continuity. An optimal plan is identified to be module for repetitive construction projects. As shown in
nondominated if no other plan can provide a better solution in Fig. 4, the scheduling module returns the computed values of
both objectives simultaneously 共i.e., a combination of shorter the project duration and total interruption days for each re-
project duration and higher level of crew work continuity兲. For source utilization plan after considering all the scheduling
example, all the generated plans in Fig. 2 are nondominated be- constraints of crew availability, crew work continuity and job
cause no other plans can be found to provide a better performance logic.
in both objectives simultaneously. Conversely, plan C is domi- 2. Perform genetic operators 共i.e., selection, crossover, and mu-
nated by all the generated plans in group D because each of these tation兲 to create a new child population. These three opera-
plans provides a combination of a shorter project duration 共less tors adopt the survival of the fittest principle to provide an
than 110.4 days兲 and a higher level of crew work continuity 共less effective evolutionary mechanism for generating improved
than 16 days of interruption兲, as shown in Fig. 2. The present resource utilization plans over a number of generations. First,
optimization module is designed to preclude all dominated solu- the selection operator favors solutions with better fitness val-
tions in order to enable the generation of only the non-dominated ues to move to the reproduction phase. Second, the crossover
optimal construction plans 共see Fig. 2兲. operator swaps a chunk of the genetic material between the
In this optimization module, each of the identified optimal mated pairs of solution in the reproduction phase. Third, the
construction plans specifies an optimal solution for two important mutation operator is used to change the genetic materials in
resource utilization variables: 共1兲 crew formation 共ni兲 that repre- the produced child population randomly to avoid conver-
sents the availability of alternative crew utilization options for gence to local optimal solutions.
activity 共i兲 with varying daily productivity rates 共Pi,n兲; and 共2兲 3. Evaluate the fitness functions 共i.e., project duration and total
crew interruption vector 共vi兲 that represents the interruption days crew interruption days兲 for each solution in the newly created
to the work continuity of crew 共ni兲 when it moves from one re- child population Ct in a similar process to that described in
petitive section 共j兲 to the next 共El-Rayes and Moselhi 2001兲. This step 共1兲.
module requires construction planners to specify all feasible crew 4. Combine child and parent populations 共Ct and Pt兲 to form
formations 共ni兲 for each activity 共i兲 in the project, however it does newly combined populations, and then select the best 50% of
not require them to provide data for the feasible crew interruption the members of the combined population to form a new
vectors 共vi兲 which are automatically generated and evaluated by parent population for the next generation. This process rep-
the module. resents a strong form of elitism as it enables preserving the

Table 1. Quantities of Work and Available Crew Formation Options

Repetitive activity 共i兲 Excavation 共i = 1兲 Foundation 共i = 2兲 Columns 共i = 3兲 Beams 共i = 4兲 Slabs 共i = 5兲

Repetitive unit 共j兲 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4

Quantity of work 1,147 1,434 994 1,529 1,032 1,077 943 898 104 86 129 100 85 92 101 80 0 138 114 145
共Qi,j兲 共m3兲
Available crew options
Crew formation 共ni兲 1 1 2 3 1 2 3 1 2 3 4 1 2
Productivity 91.75 89.77 71.81 53.86 5.73 6.88 8.03 9.9 8.49 7.07 5.66 28.73 7.76
共Pi,n兲 共m3 / day兲

JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / JANUARY 2006 / 15

Table 2. Solutions Generated by Previous Optimization Models
Russell and Caselton El-Rayes and Moselhi
Selinger model 共1980兲 model 共1988兲 model 共2001兲
Optimal crew Interruption Optimal crew Interruption Optimal crew Interruption
Activity formation days formation days formation days
共i兲 共ni쐓兲 共Interi兲 共ni쐓兲 共Interi兲 共ni쐓兲 共Interi兲
Excavation 1 0 1 0 1 0
Foundation 2 0 1 4 1 6
Columns 3 0 3 0 3 0
Beams 3 0 1 12 1 9
Slabs 1 0 1 0 1 0
Total interruptions 共R兲 0 16 15
Project duration 共D兲 117.9 110.4 106.8

best members of the parents’ population over generations mance level to the obtained minimum duration solution 共Min Dg兲,
共Deb et al. 2000兲. and a planner-specified level 共e.g., 0 or 20%兲 that reflects the
The above computation steps of 1–4 are repeated over a num- degree of satisfaction with the maximum duration solution
ber of specified generations 共t = 1 – T兲 in order to yield a Pareto 共Max Dg兲, as shown in Fig. 5. This function can then be used to
optimal set of nondominated resource utilization plans for the evaluate the performance of all intermediate project durations
repetitive construction project. Each plan in this Pareto optimal 共Dg兲 as shown in Eq. 共12兲 and Fig. 5. The performance in crew
set provides: 共1兲 an optimal construction plan that provides the work continuity 共PRg兲 can also be evaluated in a similar way
least project duration 共D*兲 that can be achieved at a given crew using Eq. 共13兲. For example, the utility values that represent
work continuity level 共R*兲, as shown in Fig. 2; and 共2兲 an optimal project duration performance levels 共PDg兲 for solutions A, B, and
level of resource utilization for each construction activity 共i兲 in E 共see Fig. 2兲 are 20, 100, and 70%, respectively; while those
the project that specifies the selected crew formation 共ni*兲 and representing their crew work continuity performance levels 共PRg兲
crew interruption vector 共vi*兲, as shown in Fig. 3. A construction are 0, 100, and 47%, respectively, as shown in Fig. 5. Both
planner can select, from this set, the best overall plan that satisfies performance levels 共PDg and PRg兲 are then combined using an
the specific project requirements, using the ranking and selection overall score 共Cg兲 that represents the combined performance of
module which is described in more details in the next section. solution 共g兲 in both objectives. This overall score 共Cg兲 can be
computed using planner-specified weights that reflect the relative
importance of project duration 共WtD兲 and crew work continuity
Ranking and Selection Module 共WtR兲 to the decision maker as shown in Eq. 共14兲

The objective of this module is to rank all the optimal solutions

共g = 1 – G兲 obtained from the previous module to facilitate the se-
PDg = P Max Dg + 冋 Max Dg − Dg
Max Dg − Min Dg
⫻ 共100 % − P Max Dg兲 册
lection of the best overall plan for executing the project. To ac-
共12兲
complish this, the present module uses the multiattribute utility

冋 册
theory 共von Winterfeld and Edwards 1986兲 to enable construction
planners to express their degree of satisfaction about the gener- Max Rg − Rg
PRg = P Max Rg + ⫻ 共100 % − P Max Rg兲
ated project durations and crew work continuity levels. For ex- Max Rg − Min Rg
ample, a linear utility function can be used to evaluate the perfor-
共13兲
mance 共PDg兲 of each generated project duration 共Dg兲, using a
performance scale that ranges from 0 to 100% 共see Fig. 5兲. This
function can be easily developed by assigning a 100% perfor- Cg = PDg ⫻ WtD + PRg ⫻ WtR 共14兲

Fig. 6. Example utility functions for project duration and crew work continuity

16 / JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / JANUARY 2006

Table 3. Optimal Solutions Generated by Present Model
Project performance Selected crew formation 共ni兲 Selected interruption vectors 共v j兲
Project Total crew Inter2,j Inter3,j Inter4,j
duration interruption
Solution in days days Excavation Foundation Columns Beams Slabs
共g兲 共D兲 共R兲 共i = 1兲 共i = 2兲 共i = 3兲 共i = 4兲 共i = 5兲 j=1 j=2 j=3 j=1 j=2 j=3 j=1 j=2 j=3
1 106.8 15 1 1 3 1 1 1 0 5 0 0 0 0 6 3
2 107.0 14 1 1 3 1 1 1 0 5 0 0 0 0 6 2
3 107.6 13 1 1 3 1 1 1 0 5 0 0 0 0 5 2
4 108.5 11 1 1 3 2 1 1 0 5 0 0 0 0 4 1
5 109.0 10 1 1 3 2 1 1 0 5 0 0 0 0 4 0
6 109.9 9 1 1 3 2 1 1 0 5 0 0 0 0 3 0
7 110.9 8 1 1 3 2 1 0 0 5 0 0 0 0 3 0
8 111.4 7 1 1 3 3 1 1 0 5 0 0 0 0 1 0
9 112.3 6 1 1 3 3 1 0 0 5 0 0 0 0 1 0
10 113.3 5 1 1 3 3 1 0 0 4 0 0 0 0 1 0
11 114.3 4 1 1 3 3 1 0 0 3 0 0 0 0 1 0
12 115.3 3 1 1 3 3 1 0 0 2 0 0 0 0 1 0
13 116.3 2 1 1 3 3 1 0 0 1 0 0 0 0 1 0
14 116.8 1 1 2 3 3 1 0 0 0 0 0 0 0 1 0
15 117.9 0 1 2 3 3 1 0 0 0 0 0 0 0 0 0
Interi,j ⫽ interruption days in activity i after repetitive unit j.

Application Example Caselton 共1988兲, and El-Rayes and Moselhi 共2001兲, producing
three different solutions as shown in Fig. 1 and Table 2. The
An example of a three-span concrete bridge is analyzed in order present model was utilized to analyze the same example in order
to illustrate the use of the present model and demonstrate its to: 共1兲 enable a comparison between its results and those gener-
capabilities in generating and evaluating optimal tradeoffs ated by available models in the literature; and 共2兲 illustrate its
between project duration and crew work continuity. The project capability of evaluating and ranking the obtained tradeoff solu-
consists of five construction activities: excavation, foundations, tions between project duration and crew work continuity.
columns, beams, and slabs that are repeated in four sections of the First, the model was used to generate a set of 15 optimal and
project, as shown in Fig. 1. The precedence relationships among nondominated construction plans, where each represents an opti-
these five successive activities are finish to start with no lag time. mal and unique tradeoff between project duration 共D兲 and crew
Table 1 summarizes the quantities of work for each activity in the work continuity 共R兲, as shown in Fig. 2 and Table 3. The results
four repetitive sections as well its available crew formation op- generated by the present model were compared to those produced
tions and their daily productivity rates. The example was previ- by previous models 共see Fig. 2 and Table 3兲. The results of this
ously analyzed in the literature by Selinger 共1980兲, Russell and comparative analysis illustrate the capabilities of the present

Table 4. Ranking of Generated Optimal Solutions

Performance in project objectives Overall score in % 共ck兲
Duration Crew Duration Work continuity Scenario 1 Scenario 2 Scenario 3
Solution in days interruptions performance performance in WtD = 80% WtD = 20% WtD = 60%
共g兲 共Dg兲 in days 共Rg兲 in % 共PDg兲 % 共PRg兲 WtR = 20% WtR = 80% WtR = 40%
1 106.8 15 1.00 0 80.2 20.0 60.1
2 107.0 14 0.98 0.067 80.0 25.0 61.7
3 107.6 13 0.94 0.133 77.9 29.5 61.7
4 108.5 11 0.88 0.267 75.7 38.9 63.5
5 109.0 10 0.84 0.333 73.8 43.5 63.7
6 109.9 9 0.77 0.400 69.9 47.5 62.4
7 110.9 8 0.71 0.467 65.8 51.4 61.0
8 111.4 7 0.67 0.533 64.3 56.1 61.5
9 112.3 6 0.60 0.600 60.1 60.0 60.1
10 113.3 5 0.53 0.667 55.7 63.9 58.4
11 114.3 4 0.46 0.733 51.3 67.8 56.8
12 115.3 3 0.39 0.800 46.8 71.7 55.1
13 116.3 2 0.31 0.867 42.4 75.6 53.5
14 116.8 1 0.28 0.933 41.0 80.3 54.1
15 117.9 0 0.20 1.000 36.0 84.0 52.0

JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / JANUARY 2006 / 17

model in: 共1兲 generating identical optimal solutions to those generated tradeoff solutions to facilitate the selection of the best
produced by the two existing models of Selinger 共1980兲 and overall plan for constructing the project. These new and unique
El-Rayes and Moselhi 共2001兲, as shown in solutions A and B in capabilities should prove useful to construction planners and is
Fig. 2; 共2兲 producing a wide range of optimal tradeoff solutions expected to advance existing planning and scheduling practices
that cannot be developed using existing models, as shown in for repetitive construction projects.
Fig. 2; and 共3兲 obtaining optimal tradeoff solutions that dominate
and provide superior performance in both optimization objectives
simultaneously than that obtained by the Russell and Caselton Acknowledgment
model 共1988兲, as highlighted by group D and solution C in Fig. 2,
respectively. The writers gratefully acknowledge the financial support provided
Second, the present model was used to evaluate and rank the by the National Science Foundation for this research project
15 generated tradeoff solutions to facilitate the selection and ex- under NSF CAREER Award No. CMS-0238470.
ecution of the best overall plan for the project. As stated earlier,
construction planners need to specify: 共1兲 their degree of satisfac-
tion with the maximum duration solution and the maximum Notation
interruption solution, as shown in Fig. 5; and 共2兲 the relative
importance of minimizing the project duration and maximizing The following symbols are used in this paper:
crew work continuity in the form of two relative weights Cg ⫽ overall score that represents combined
共WtD and WtR兲. In this example, it is assumed that the planner performance of solution g in both objectives;
expressed a satisfaction level of 共P Max Dg = 20% 兲 and D ⫽ project duration;
共P Max Rg = 0 % 兲 with the maximum duration and the maximum Dg ⫽ project duration of solution g;
interruption solutions, respectively, as shown in Fig. 5. Further- di,j ⫽ construction duration of activity i in repetitive
more, three different combinations of weights are evaluated in unit j;
this example to illustrate the impact of the weighting process on FCrew关i,j兴 ⫽ earliest finish times for activity i in unit j that
the ranking of the tradeoff solutions, as shown in Table 4. satisfy crew availability and crew work
As expected, the first 共WtD = 80% and WtR = 20%兲 and second continuity constraints;
共WtD = 20% and WtR = 80%兲 combinations of weights that favored Fi,j ⫽ scheduled finish time for activity i in
the performance in project duration and crew work continuity repetitive unit j;
produced the highest overall score for the least duration solution Interij ⫽ interruption days in activity i after repetitive
共g = 1兲 and the least interruption solution 共g = 15兲, respectively unit j;
共see Table 4兲. On the other hand, the third combination of weights Max Dg ⫽ maximum duration solution;
共WtD = 60% and WtR = 40%兲 produced an intermediate tradeoff Max Rg ⫽ maximum interruption solution;
solution 共g = 5兲 as shown in Table 4. This illustrates the capability Min Dg ⫽ minimum duration solution;
of the present model in performing instant sensitivity analysis in Min Rg ⫽ minimum interruption solution;
order to study the effects of varying the weights on the ranked Pi,n ⫽ productivity rate of crew n in activity i;
solutions, without the need to rerun the computationally demand- PDg ⫽ project duration performance level;
ing genetic algorithms each time these weights are changed, as PRg ⫽ crew work continuity performance level;
shown in Table 4. Qi,j ⫽ quantity of work of activity i in repetitive
unit j;
R ⫽ total number of crew interruption days;
Conclusions Rg ⫽ total interruption days in solution g;
SCrew关i,j兴 ⫽ earliest start time for activity i in unit j that
A robust multiobjective optimization model was developed to satisfy crew availability and crew work
support the planning and scheduling of repetitive construction continuity constraints;
projects. The model enables construction planners to generate and Si,j ⫽ scheduled start time for activity i in repetitive
evaluate optimal construction plans that establish optimal unit j; and
tradeoffs between project duration and crew work continuity. SLogic关i,j兴 ⫽ earliest start time that satisfies job logic and
Each of these plans identifies, from a set of feasible alternatives, precedence relationships between activity i
an optimal level of resource utilization for each activity in the and all its predecessors.
project. To accomplish this, the model incorporates: 共1兲 a sched-
Subscripts and Superscripts
uling module that calculates the project duration and level of crew
work continuity for repetitive construction projects; 共2兲 a multi- G ⫽ number of nondominated optimal solutions
objective optimization module that searches for and identifies 共g = 1 – G兲;
near optimal construction plans; and 共3兲 a ranking and selection I ⫽ number of construction activities 共i = 1 – I兲;
module that ranks the generated solutions according to the spe- J ⫽ number of repetitive units 共from j = 1–J兲.
cific requirements of the project being considered. An application K ⫽ population size 共k = 1 – K兲; and
example was analyzed to enable a comparison between the T ⫽ number of generations 共t = 1 – T兲.
present model and those available in the literature. The results of
this analysis illustrates the new capabilities of the developed
model in: 共1兲 generating all optimal tradeoff solutions between References
project duration and crew work continuity in a single run, where
each provides the least project duration that can be achieved at a Ashley, D. 共1980兲. “Simulation of repetitive-unit construction.” J. Constr.
given level of crew work continuity; and 共2兲 ranking these Div., Am. Soc. Civ. Eng., 106共2兲, 185–194.

18 / JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / JANUARY 2006

Coello, C., and Lechuga, M. 共2001兲. “MOPSO: A proposal for multiple Harris, B., and Ioannou, P. 共1998兲. “Scheduling projects with repeating
objective particle swarm optimization.” Technical Rep. No. activities.” J. Constr. Eng. Manage., 124共4兲, 269–278.
EVOCINV-01-2001.CINVESTAV, Instituto Politecnico Nacional. Col. Hegazi, T., and Wassef, N. 共2001兲. “Cost optimization in projects with
San Pedro, Zacatenco, Mexico. repetitive nonserial activities.” J. Constr. Eng. Manage., 127共3兲,
Deb, K., Pratap, A., Agrawal, S., and Meyarivan, T. 共2000兲. “A fast elitist 183–191.
nondominated sorting genetic algorithm for multiobjective optimiza- Kandil, A., and El-Rayes, K. 共2005兲. “Parallel computing framework for
tion: NSGA-II.” Proc., 6th Parallel Problem Solving from Nature VI optimizing construction planning in large-scale projects.” J. Comput.
Civ. Eng. 19共3兲, 304–312.
Conf., Paris, 849–858.
Moselhi, O., and El-Rayes, K. 共1993兲. “Scheduling of repetitive projects
El-Rayes, K. 共2001兲. “Optimum planning of highway construction under
with cost optimization.” J. Constr. Eng. Manage., 119共4兲, 681–697.
A ⫹ B bidding method.” J. Constr. Eng. Manage., 127共4兲, 261–269. Russell, A., and Caselton, W. 共1988兲. “Extensions to linear scheduling
El-Rayes, K., and Kandil, A. 共2005兲. “Time-cost-quality trade-off analy- optimization.” J. Constr. Eng. Manage., 114共1兲, 36–52.
sis for highway construction.” J. Constr. Eng. Manage., 131共4兲, Selinger, S. 共1980兲. “Construction planning for linear projects.” J. Constr.
477–486. Div., Am. Soc. Civ. Eng., 106共2兲, 195–205.
El-Rayes, K., and Khalafallah, A. 共2005兲. “Tradeoff between safety and Senouci, A., and Eldin, N. 共1996兲. “Dynamic programming approach to
cost in planning construction site layouts.” J. Constr. Eng. Manage., scheduling of nonserial linear project.” J. Comput. Civ. Eng., 10共2兲,
共in press兲. 106–114. 2, April 1996.
El Rayes, K., and Moselhi, O. 共1998兲. “Resource-driven scheduling of von Winterfeld, D., and Edwards, W. 共1986兲. Decision analysis and be-
repetitive activities.” J. Constr. Manage. Econom., 16共4兲, 433–446. havioral research, Cambridge University Press, Cambridge, England.
El-Rayes, K., and Moselhi, O. 共2001兲. “Optimizing resource utilization Zitzler, E., and Thiele, L. 共1999兲. “Multiobjective evolutionary algo-
for repetitive construction projects.” J. Constr. Eng. Manage., 127共1兲, rithms: A comparative case study and the strength Pareto approach.”
18–27. IEEE Trans. Evolutionary Computation, 3共4兲, 257–271.

JOURNAL OF MANAGEMENT IN ENGINEERING © ASCE / JANUARY 2006 / 19

View publication stats